The Consequence of Feedback I: Coevolution
The Consequence of Feedback I: Coevolution
Abstract and Keywords
Feedback is a twentieth-century name for a dynamic that has been understood to be basic to evolution at least since Darwin’s proposal of natural selection. Feedback is an important aspect of the so-called “modern evolutionary synthesis,” which guided thinking through much of the twentieth century and led especially to gene-centered views. And it has loomed particularly large in the “extended evolutionary synthesis” that has taken shape over the last thirty years and broadened such gene-centrism to take in other aspects of evolution. Feedback is paramount in coevolution, the processes by which organisms’ adaptational histories shift in relation to those of other organisms, both across different species and within the same species. Positive and negative feedback are each active in coevolution, but the effects of positive feedback are especially rich and, in some instances, unexpected. Standing outside both of these kinds of feedback is another aspect of evolutionary systems, one that has not received sufficient attention: feedforward, the impact of elements that influence or control feedback cycles but are not changed by them.
: 1 : Feedback and evolution
Feedback is hard to think. The idea that A causes B which causes A looks simple, but its circularity breeds confusion, flying in the face of deep-seated conditions governing how we form knowledge about the world. The challenge of thinking even a simple feedback system confronts us in everyday life. When we turn on our furnace, we understand ourselves to be causing it to start up, but the thermostat mechanism by which it is turned off without our assistance once the house is heated is a bit less easy for us to take in. The compounding of interest on credit-card debt is trickier still to grasp—a fact banks exploit to submerge millions of customers in long-lasting debt. Einstein is supposed to have called compound interest “the most powerful force in the universe”; with less hyperbole he added (the story continues), “He who understands it, earns it; he who doesn’t, pays it.”
The confusion bred by feedback systems arises also in less quotidian pursuits and helps to explain the historical fact that feedback theories came late, long after the design of feedback machines themselves. These machines have a history reaching back to the ancient world, where water clocks were equipped with float valves that maintained a constant water level by closing off input as the level rose and opening it as it fell. A more recent feedback mechanism, the centrifugal governor that James Watt designed in 1788, was also used for regulation, in this case of the new and epoch-making steam engine. Here power from the engine spun a system of fly-balls; as they spun faster and rose, they gradually closed a valve regulating steam flow into the engine, damping its performance. Watt’s device and similar mechanisms that followed stimulated early moves to generalize their operation, most famously in James Clerk Maxwell’s paper “On Governors,” delivered to the Royal Society of London in 1868. This represents (p.20) an important step toward modern control theory (Mayr 1971), and indeed, the conceptualizing of feedback only truly came of age in mid-twentieth-century control, or “cybernetic,” theory, dynamic systems theory, and information theory (Wiener 1948; Bertalanffy 1969). These bodies of thought all emerged in the wake of additional technology: the design of electronic circuits and exploration of the practical consequences of “feeding” the output of a circuit “back” into it. Finally, in the late twentieth century, theory and technology were fully joined in digital systems and processing, in a development that has come to be of such pervasive importance that we might justifiably call our time the age of feedback.
The difficulty of conceptualizing feedback is also plain to see in biological theory. Here feedback systems play a fundamental role, and today they are identified or postulated on every scale: from molecular networks governing the expression of genes and cell metabolism, through the life cycles of individual organisms and species, all the way to epochal processes and major transitions in the history of life. But the role of feedback remains hard to pin down, hard to model, and hard to formalize. The situation is clearest in the area of evolutionary theory (Robertson 1991). Feedback operations are basic to ideas that have joined together over the last several decades to form what is sometimes called the “extended evolutionary synthesis,” an outgrowth of the “modern synthesis” that had united Darwinian natural selection and Mendelian population genetics around 1930 (Pigliucci and Müller 2010). The novel ideas making up the extended synthesis remain controversial, however, and this resistance has been traced in part to the models of “reciprocal causation” that their feedback operations introduce (Laland et al. 2015). It is revealing that the major study of one of these ideas, niche construction, dates from as late as 2003—and even at that date could justly be subtitled The Neglected Process in Evolution (Odling-Smee, Laland, and Feldman 2003).
Darwin and feedback
There is an irony in this situation, because feedback mechanisms were basic to the hypotheses of evolutionary theory from its start and are prominently featured in Darwin’s On the Origin of Species of 1859. The basic dynamic of natural selection itself shows elements of feedback design. Organisms inherit their traits from previous generations but also vary them, and this variety is either culled by environmental constraints or thrives in them. If it thrives, environmental influence in effect feeds back onto a population so as to increase the frequency of the trait in successive generations, and this, under circumstances Darwin spent much of his book proposing, might eventually lead to new species. An even clearer feedback system in the Origin involves the interplay of one species with another. The grand “entangled bank” with which Darwin ended his book was the result (p.21) of a “struggle for existence” whose mutual balances were determined by natural selection. Darwin envisaged a give-and-take among the different organisms in any ecosystem in which, “if some of these many species become modified and improved, others will have to be improved in a corresponding degree or they will be exterminated” ( 2003, 620). Today this mutually adaptive entanglement of species travels under the name coevolution.
If we look to the background of Darwin’s vision, we find another feedback system. His inspiration in 1838 for the struggle for existence, and the starting point for his exposition of it later on in the Origin, was Thomas Malthus’s An Essay on the Principle of Population of 1798 (Malthus 2008). Malthusian population dynamics is itself a feedback design, built on the mutual interaction of population size and resource availability. In summary form: A population increases faster than the availability of its resources, resources per capita decline, the underresourced population declines, and resources rise again. A changes B which changes A which changes B. And the connection of natural selection to feedback mechanisms was more than a buried implication in the nineteenth century. In the now-famous paper he sent to Darwin the year before the publication of the Origin, Alfred Russel Wallace, the codiscoverer of the theory, hypothesized a mechanism by which the varieties of living things must diverge ever farther from their original types. Wallace’s mechanism ensured a balance of organisms’ capacities with their needs and the resources of their habitats, and in this, he wrote, it operated exactly like Watt’s self-regulating centrifugal governor (Wallace 1858).
Feedback is nothing new to evolutionary theory, then. That it characterizes the basic dynamics of both adaptation and population flux was understood by their first theorists. Today, however, the new ideas of the extended synthesis of evolutionary theory have expanded and enriched the roles of feedback as never before. It is the task of this chapter and the next to review and in some measure elaborate these theories as a foundation on which to build a new view of human evolution.
: 2 : The modern synthesis
The modern synthesis of evolutionary theory took shape across the first four decades of the twentieth century and merged the two greatest achievements of nineteenth-century biology: Gregor Mendel’s genetics and Darwin’s natural selection. It can come as a surprise to learn that these achievements, so fundamental now, languished for decades after their first formulations. Mendel’s analysis of pea plants in the years around 1860 showed that the traits of individual plants could be tracked through generations as the interplay of discrete, particulate units of inheritance; (p.22) but his findings were largely ignored until 1900. Natural selection, on the other hand, was certainly not ignored, but it suffered in a different way: exactly by virtue of its proliferation. Darwin’s Malthusian struggle for existence was one of those ideas irresistible in its moment, and it came, across the decades after 1859, to be dispersed into all areas of popular doctrine, providing an ideological umbrella under which ideas scientific and social, reputable and reprehensible, could gather. The proliferation of the idea was accompanied by a dilution of the power of Darwin’s specifically biological conception, a dilution perhaps inevitable as long as there was no discernible unit of inheritance and variation.
The scientific rehabilitation of both ideas began when they were joined together in the years after 1900; this was the beginning of the modern synthesis (Huxley  2010; Depew and Weber 1997). Mendel had linked the traits of individual organisms (by now called their phenotypes) to their particulate units of inheritance (their genes, collectively making up their genotypes). His finding was now extrapolated mathematically to whole populations of organisms, in analyses of the varying frequencies in populations of distinct phenotypes and the genotypes they “expressed.” Particulate inheritance met statistical analysis, and population genetics was born. The analysis of change in gene frequencies across generations immediately suggested the relevance of the new field to evolutionary theory, and Darwin’s natural selection came into a new, sharp focus. The three ingredients of his algorithm had always been clear: inheritance of traits, variation of traits, and unequal selection of the variants. Now it was understood that variations in genotypes (mostly through mutations) offered up new individual phenotypes to environmental constraints and that, as these variants were differentially selected, the frequency of traits in the whole population would change. Here, finally, the motor of evolution through natural selection came into view, the genetic means by which populations of organisms could change over time: varying, thriving, dying off, and—most important—diverging into new species.
This synthesis of natural selection theory and population genetics came into its first maturity in the years between World Wars I and II. From the 1940s on through the next several decades, its powerful implications were explored by the major figures of genetic and evolutionary biology. These are the years of giants in the field: Ronald Fisher, J. B. S. Haldane, and Sewall Wright at the beginning of the period, Theodosius Dobzhansky, Ernst Mayr, George Gaylord Simpson, and John Maynard Smith later. Along the way, in 1953, the synthesis received the most dramatic kind of support with the discovery of the structure of the genetic molecule, DNA, by James Watson, Francis Crick, Rosalind Franklin, and others. If Mendel’s work had revealed (p.23) the motor of natural selection, now its operation was known in new detail, and molecular biology leapt forward with a momentum that has not slackened today.
Selfish genes and memes
This momentum brought a crescendo of acclamation for the cen-trality of the gene in evolution: the so-called “gene-centered” view. The loudest support came from Richard Dawkins, who in the 1970s offered an account of genes as “selfish” replicators that use the bodies of the organisms they inhabit as “vehicles” or “survival machines” and exploit metabolic energies and resources for their own propagation (Dawkins 1976). Even the suite of an organism’s actions in the world—what Dawkins (1982) would call its “extended phenotype,” as seen, for example, in a termite mound or beaver dam—comprised first and foremost mechanisms ensuring the survival and replication of the genes controlling the actions.
Pushback against Dawkins’s extreme position was quick to form, voiced by Richard Lewontin, Stephen Jay Gould, and others (Lewontin 1983). They did not dispute in any general way the consequence of genes in evolution but aimed to modify Dawkins’s position by asserting the consequence of phenotype and environment in addition to genes and by broaching the complexities of interaction among all three (Lewontin 2000). It is not too much to say that all the major moves in the extension of the modern synthesis of evolutionary theory have concerned these interactions in their various forms and at various levels of scale, from molecular to ecosystemic. And, since the potential for feedback relations in these interactions is a rich one, the debates over the gene-centered view prepared the ground for advances in this area also.
One other aspect of Dawkins’s gene-centrism needs to be taken up here, since we are concerned with the quintessential cultural animal, Homo sapiens. Dawkins carried the model of gene action over by analogy to human culture. He proposed that culture has its own replicators, analogous to genes in their particulate nature, their transmission, and the action on them of selection. He called these replicators memes (Dawkins 1976)—and memes, like Darwin’s struggle for existence a century before, proved to be an idea irresistible to popular culture and popular science alike. The term spread quickly through cultural analysis, giving rise to the ersatz field of “memetics” and becoming the term of art to signal every recurrence of a cultural gesture, every sign of a cultural connection, and finally almost any cultural move that we recognize: the Nike swoosh, the Coke C, connections between two hit songs, the repeating themes of a political season.
But if the word is by now lodged in all manner of public parlance, the concept remains problematic. It is important to identify exactly where the difficulty lies. It is not found in memes’ signaling of the transmissible nature (p.24) of culture. This is a given of all human culture, which, complicated and hard to analyze as it is, does not need memetics to emphasize it. Neither is the problem the idea that cultural selection might take place along lines analogous to natural selection. Dawkins borrowed this insight from population geneticist Luigi Luca Cavalli-Sforza, and, as we will see in chapter 3, it has stimulated well-developed positions in the extended evolutionary synthesis. The problem with the meme concept is not even the idea that discrete, exactly duplicated gestures play an important role in culture. In itself, this idea is unassailable, as there is and always has been plenty of more or less exact duplication in human cultures. Those swooshes and Cs may exemplify modern, precise mechanical and electronic reproduction, but we can trace their kind of gestural repetition all the way back to the design resemblances of Lower Paleolithic hand axes a million years ago.
The problem of memes concerns, instead, an assumption that is built on all three of these premises: the notion that culture evolves according to the Darwinian algorithm of selection because of the particulate nature it shares with genes. At a time when even the particulate nature of genes has been blurred from various quarters of developmental and molecular biology, memetics advances an idea of cultural development and history as a flow of discernible, delimitable units. Every attempt to define these, however, runs up against the proliferative, nonparticulate tendencies of human sociality in all its modes (Bloch 2000). The Nike swoosh may be repeatable, but it is fundamentally unlike a gene for a pink petal on a pea plant, since it is no singular gesture at all but always both the product of and an agent in a changing network of cultural meanings. And it is the place of a duplicated cultural gesture in the network that is crucial in making it meaningful. (For Terrence Deacon  this reveals memetics to be merely a schematic simplification of the study of the systems of signs that found all cultures.) Genes, however complex and elusive their identities, transmit their effects though a conservative system in which nucleotide sequences are copied faithfully in the vast majority of their replications, miscopied rarely. Cultural gestures, even those ostensibly reproduced exactly, enter onto a terrain of meaning where they differ in each new instance—even while looking identical to the memeticist, even in those cultural settings that the anthropologist or sociologist might call “conservative.” To isolate a meme’s duplication, then—the chief aim of most memeticists thus far—is exactly to render it meaningless, to cancel out all the meanings that bubble up, fade away, and change with its repetition, and therefore to set aside all the reasons why we might wish to understand its duplication. Cognitivist and anthropologist Dan Sperber, long an opponent of memetics and the notion of culture as replication, has summed it up this way: “A process of (p.25) communication is basically one of transformation” (1996, 83). This is a truth about human culture that is not limited to its modern forms but reaches back dozens of millennia, at least.
That memes ultimately offer a reductivist view of human culture, its transmission, and its evolution is recognized today even by those thoughtful students of cultural evolution who wish to retain Dawkins’s concept in qualified form (e.g., Henrich, Boyd, and Richerson 2008). Memes point the way not to a cultural science but to a cultural scientism, a method for generating thin descriptions; and in their reductivism they are akin to a problem that we will identify in other, more serious scientific approaches to culture. The alternative developed in this book is the amalgam of science, humanistic theory, and historicism introduced in chapter 1.
: 3 : Coevolution
In the wake of the gene-centered view came the extended synthe-sis in evolutionary theory, with its expanded roles for feedback mechanisms. The first aspect of it I will take up is coevolution: reciprocal relations between the evolutionary adaptations of different species. Darwin had already posited this in On the Origin of Species—for example, in this meditation on clovers and bees:
The tubes of the corollas of the common red and incarnate clovers … do not on a hasty glance appear to differ in length; yet the hive-bee can easily suck the nectar out of the incarnate clover, but not out of the common red clover, which is visited by humble-bees alone. … Thus it might be a great advantage to the hive-bee to have a slightly longer or differently constructed proboscis. On the other hand, I have found by experiment that the fertility of clover greatly depends on bees visiting and moving parts of the corolla. … Hence, again, if humble-bees were to become rare in any country, it might be a great advantage to the red clover to have a shorter or more deeply divided tube to its corolla, so that the hive-bee could visit its flowers. Thus I can understand how a flower and a bee might slowly become, either simultaneously or one after the other, modified and adapted in the most perfect manner to each other, by the continued preservation of individuals presenting mutual and slightly favourable deviations of structure. ( 2003, 611)
The feedback pattern in Darwin’s example is clear. It is advantageous for clovers to be pollinated, advantageous for bees to suck clover nectar. Natural selection will therefore favor bee variants with enhanced ability to gain (p.26) access to the nectar, and clover variants making nectar more accessible to the pollinators. Through two separate vectors of selection, two species come to be linked as determining elements in each other’s selective environments. If there is a perturbation in the local ecology—here, the disappearance of the humble-bees—the selective pressures on the clovers and the remaining hive-bees will be adjusted. The mutual adaptation of the species, their coevolution, will continue in a reshaped way.
In its classic usage, the term coevolution refers to this interlinked, feedback relation of the selective histories of species, groups, or populations. These can be, as here, histories of mutual advantage, but they need not be. The relations of predator species and the species they prey upon form coevolutionary pairs. As enhanced modes of attack (or defense) are selected in one species, variants in the other species are selected that alter it toward more effective modes of defense (or attack). This is the well-known evolutionary “arms race.” Darwin saw this possibility too and described it in a hypothetical example of wolves and deer that came just before the clover-and-bee example in the Origin.
Symbiotic relations of all kinds also involve coevolutionary feedback. The impact of such symbiotic coevolution has, by all recent accounts, been immense in the history of life. To cite two major, early episodes of it, now generally accepted: both the mitochondria found today in most complex cells and the chloroplasts found in photosynthetic cells are the remnants of the invasion of ancient cells without such organelles by other, single-celled symbionts. The subsequent coevolutionary history—by which, for example, the DNA, RNA, and proteins of the host cells came to be linked metabolically to those of the eventual mitochondria—must in each case have been long and complex.
Optima networks; the Red Queen
Of course, groups of organisms in the real world, whether species or other taxonomic levels, are not linked to one another in a one-to-one, exclusive relation. Ecosystems are entanglements (to recall Darwin’s word) of countless groups, mutually bound in beneficial or harmful ways, and the connections of their selective histories extend promiscuously in many directions. For any individual population of related organisms we can picture the lines of coevolutionary connection as forming a network of selective relations with other groups, a network that extends and changes across evolutionary time. Bees compete for nectar with other insects and birds, and they fight off microorganismal parasites, predators for which they are prey. Clovers vie with many other plants around them for resources in the soil or space in the sun; they must survive being grazed by passing herbivores and might depend on those animals for dispersal to new terrains. All these relations are not static but change with each perturbation (p.27) of the ecosystem: a new herbivore, a more aggressive microorganism, a fungus that kills off a rival plant.
In the large scheme of any ecosystemic entanglements, most coevolutionary relations will not be mutually beneficial in the happy manner of clovers and bees. This seems intuitively right, since all the organisms in an ecosystem are competing to enlarge their share of the limited energy bound up in it. Darwin’s struggle for existence is on the whole truly a struggle, not a coalition for mutual advancement. In 1973 the evolutionist Leigh Van Valen gauged this struggle against the paleontological evidence of extinction rates for many groups. He was puzzled that, in successions of evolving life-forms, later ones did not persist longer than earlier ones, as we might expect if the successions reflected adaptations making for ever-improved fitness in the relevant “adaptive zones” or ecosystems. The mistake, he saw, was to think of the zones as static, when in reality they are “resource spaces” constantly depleted by the groups struggling to survive in them. In any adaptive zone, then, each group is faced with a similar constant depletion of the resources to be won, and its position cannot be secure. The ecosystem as a whole is “an ensemble of mutually incompatible optima,” each one the optimum for a given species or group, and each group struggles to gain and maintain its own optimum. Natural selection in this view does not result in a constant improvement of organisms in relation to their ecosystem but is more like a constant movement not to slip away from an optimum on an ever-mobile terrain. Van Valen (1973) pictured each group running to maintain its place, and, with a bow to Lewis Carroll, he dubbed his model the “Red Queen’s Hypothesis.”
In this assemblage of competing optima, feedback relations remain paramount. They mutually shape selective pressures in the many species or groups bound in an ecosystem, favoring this phenotype over that in each group and thereby gradually altering its identity. But the alterations only keep groups in the running with the alterations of all the other groups connected to them. Coevolution is nothing but this network of mutual change in interacting populations, and feedback relations extending across many generations have proved to be a fertile way of modeling ecosystems and species histories.
Levels of coevolution
I have been circumspect about specifying the exact level in tax-onomy at which the feedback operates. Is it the species level? A higher taxon such as genus or tribe? A lower one such as subspecies or variety? My caution is due to the fact that the relations must be conceptualized as working at various levels; different feedback systems will come in and out of focus according to the more or less fine grain of our investigation. To return to Darwin’s example, we can understand as he does the feedback (p.28) operating between hive-bees and red clover, or we can think about its operation between two types of bees and two types of clover. At this level too feedback systems extend across evolutionary history.
This fluidity of levels signals a broader, finally pervasive operation of feedback systems in biological organization. In recent biological theory, feedback is discerned at all levels in the organization of life and its history. For example, autocatalytic systems of molecules—molecule A catalyzes the production of molecule B which catalyzes C which in turn catalyzes A—are proposed as fundamental to the origin of life (Kauffman 1993; Maynard Smith and Szathmáry 1995). The conservation of basic body designs and features across huge stretches of evolutionary time is traced to self-maintaining networks of genes and proteins, stable feedback systems which in turn stabilize the unfolding development of organisms and resist radical alterations in it (Wagner 2014). At the opposite end of the organizational scale, ecosystems cannot be well understood except as flows of matter and energy through feedback “hypercycles”—seen, at the most basic level, in food-chain cycles in which dead matter provides sustenance for organisms successively higher on the trophic scale, which die and reenter the cycle. These ecological feedback cycles operate on all levels, from the local to the planetary (Wilkinson 2006).
Even if we narrow our focus back to the coevolutionary relations of two species only, feedback reveals multilevel complexity. Across generations a predator and its prey are linked in a coevolutionary relation, but what kinds of features determine which individuals in the predator population will be most successful in hunting or which in the prey population will be most vulnerable? Several features come quickly to mind, and some of these will be reflections of ontogeny, not a selective, phylogenetic history, as when very young or old individuals in herds of herbivores are targeted by carnivores. However, in many animal species the foremost general phenotypic distinctions are sexual traits, and these differences too can be linked to the feedback arms race. The peacock’s tail is an extraordinary display to attract the peahen, but it also requires matter and energy for its development and slows down its bearer, making him more vulnerable to predators. It is selectively advantageous only so long as the resources it requires and the disadvantages it presents do not outweigh the preferential reproductive access it gains. This means that the interspecies feedback (predator and prey) has inserted itself into intraspecies (sexual) dynamics.
Moreover, the sexual dynamic of the peacocks and peahens in itself could have arisen only through another selective feedback process—this time in the subprocess in natural selection that Darwin ( 1981) described in his second great book: sexual selection (see also Prum 2017). The magnificence (p.29) of the peacock’s tail and the sexual displays and behaviors of many other species seem so natural that we sometimes forget their own selective histories, which involve an arms race at a level different from predator and prey: the level of competition among individuals of the sex doing the displaying (usually male, but not always). To speak only of birds: bright plumage, elaborate songs, and spectacular mating dances are all costly behaviors, and the cost must be balanced by a gain calculated across evolutionary history. The calculation has to do with swaying the nondisplaying sex, of course. Gaudy tails do not attract females by some universal rule and as if by magic; there needed to be a reciprocal selection between the male and female birds that selected not only grander tails but also females that noticed the grandeur.
The history of coevolution, in this way, turns out to be one of loops networked with other loops at each level and intersecting with the loops of other levels. In other words, feedback all the way down and all the way up.
: 4 : Positive and negative feedback
This survey of coevolution puts us in position to say something more about the general nature of feedback. Feedback systems come in two varieties, negative and positive. In negative feedback systems, the output that is fed back into the circuit opposes what generates it; this is the variety that is represented by Watt’s governor (increased engine activity throttles down steam input) or that turns off your furnace once the ambient temperature reaches a certain level (here there is a threshold at which the rising room temperature, gauged in your thermostat’s thermometer, breaks the electrical circuit and turns off the furnace). Negative feedback can bring about overall stability in a system—homeostasis, in other words—if the contrary effects are brought into balance. Homeostasis will also often involve oscillations, as happens with your furnace: once the room temperature falls back below the threshold, the thermostat closes the circuit and turns on the furnace.
The balances of the natural world, from bacterial colonies to the largest and most complex ecosystems, pervasively involve negative feedback. Not only do bacteria grown on agar come up against resource limitations as the population increases, but their colonies also conform to spatial limitations determined by the diffusion of resources through them and even produce their own metabolic inhibitors of growth (Cooper, Dean, and Hinshelwood 1968; Hochberg and Folkman 1972). Among animals, a predator population is limited by the evolved defenses that make its prey more elusive, and the prey population is limited by the effectiveness of predator (p.30) hunting (as well as by other factors, such as the availability of its own foodstuff). Malthus’s population dynamics, so important for Darwin, involve a negative-feedback circuit whereby population growth is slowed by resource depletion. In taking over this dynamic as the basis for natural selection, Darwin assumed along with Malthus that any population would expand exponentially in a situation of bountiful resources. Since such a situation could not last for long as a population grows, he reasoned, natural selection would operate to weed out the most vulnerable variants in a population—functioning, in other words, as negative feedback.
If negative feedback often helps to bring about the balances of natural systems, positive feedback pushes against its constraint. The very tendency of populations to increase geometrically is an effect of positive feedback, as the population in one generation reproduces more than its number, giving rise to a population whose enlarged numbers become reproducers themselves. The explosive growth that results is not always immediately dampened in local ecosystems, as anyone who has had the flu or a strep throat can attest. Moreover, a negative feedback without any amplification in the system would soon reduce its output to nothing, so in biological systems positive feedback is regularly linked with negative feedback in oscillating cycles. Thus, Darwinian selection, with its negative feedback, is driven by the tendency to population growth. In the Malthusian model, if the human population declines because of the negative impact of reduced resources per capita—say, rabbits and birds to hunt—the rabbits and birds will rebound through the positive cycles of their own population growth. Then the human population will switch back to the positive phase of the oscillation until it is dampened again.
Positive feedback can lead to exponential, even system-disrupting growth. The concept of such “runaway” feedback is much with us these days, in everything from climatic greenhouse effects to cascading economic trends and our understanding of the geological histories of Mars or Venus. In local biological networks, runaway positive feedback can bring about dramatic outcomes if it is not braked—think of that flu or strep or of an algal bloom in a pond where runoff fertilizer phosphates have lifted constraints on nutrients.
Effects of positive feedback
In evolutionary dynamics the possible consequences of positive feedback have only begun to be explored, but it is clear that they can be dramatic and unexpected. In the 1990s geologist Douglas S. Robertson and biologist Michael C. Grant offered several scenarios (Robertson 1991; Robertson and Grant 1996). They started by using a positive-feedback model to explore the mysterious, much-observed tendency of individual species to evolve toward larger body size (a tendency (p.31) known as Cope’s Rule). Here is how their model worked: In any particular species, given loose genetic determinants of body dimensions, there will be a range of body sizes, with mid-range sizes more frequent than outliers, such that the distribution, if graphed, would form a bell curve. The peak of the curve, the most frequent body size, represents also a fitness peak, the selected body size that confers greatest fitness for the species in its environment, perhaps because it enables the greatest hunting prowess or escape from hunters, perhaps because it facilitates the best diffusion of resources through the system, or for many other reasons. At the same time, an intraspecies fitness pressure might also exist, conferring some advantage on slightly larger individuals—for example, in an animal population, an advantage in competing for mates. The second dynamic would then exert a positive feedback pushing the size distribution of the whole population away from the greatest overall fitness. The average body size of the species would grow larger, and its peak on the size distribution graph would shift, no longer coinciding with that of the fitness graph.
In addition to offering an explanation for Cope’s Rule, this model has implications concerning the evolutionary consequences of positive feedback. It suggests that sometimes the feedback can lead toward a breaking apart of a single population, with (in graphic terms) ideal sizes now clustering in two peaks, each corresponding to a different fitness peak determined by different criteria—a mechanism that might eventually result in division of the populations and even full speciation. More dramatically, as the feedback militating for larger bodies worked against overall fitness it might, in an extreme form, lead a species to shift far from its ideal fitness, toward a fitness level compromised enough to bring about its extinction—selection for unfitness! Projected onto a large, multispecies canvas, Robertson and Grant suggested, such a mechanism might explain the regular, oscillatory rhythm of extinctions in groups of species (that is, whole genera) that paleontologists have observed and wondered about.
The mechanism also suggests how a species could last for long periods in a slowly changing form, with negative and positive feedback more or less in balance, only then to reach a threshold where positive feedback became dominant and the species underwent quick transformation. This would conform to Stephen Jay Gould’s “punctuated equilibrium” model of phylogenetic history, in which long periods of evolutionary stasis are interrupted by dramatic change. This kind of dynamic might even explain quick bursts of evolutionary diversification, such as that preserved in the Cambrian deposits of the Burgess Shale (Gould 1989). But at the same time, Robertson and Grant showed, an opposed dynamic decreasing diversification could be another possible outcome of positive feedback. This could occur if its (p.32) nonlinear, exponential effects overwhelmed other forces or constraints at work in a system. In this case the feedback would “lock in” certain evolutionary and morphological tendencies, enabling them to dominate the selective landscape and in effect pointing natural selection in one direction rather than another. Such locking might explain the evolutionary history that seems to have followed the Cambrian explosion of the Burgess Shale period, which witnessed a rapid narrowing of the diversity of floral and faunal body types.
Robertson and Grant make a strong case for the pervasiveness of feedback in the evolution of life. They go so far as to see in it a fourth element in the Darwinian algorithm: inheritance, variation, selection, and feedback. This seems to me a categorical misstep, for we need to understand feedback not as an additional element so much as the systemic relation of the first two elements as they unfold in time and across generations to bring about the third. Misstep or not, however, their assertion is no exaggeration of the paramount role itself of feedback. Compound interest may not be the most powerful force in the universe, but in the history of life its effects have been large.
There is one more refinement that needs to be introduced here, although it is not usually included in discussions of evolutionary feedback. Perhaps this is because it is not technically part of a feedback dynamic at all but defines an element acting on that dynamic from the outside.
With inevitable simplification, we can think of biological feedback systems as closed, insofar as the feedback itself concerns the specific traits that form the loop. The linked selective histories of clover and bee or peacock and peahen, for example, form closed cycles. In the first case, ease of access to nectar, through traits selected in both species, closes the loop; in the second, the female’s registering of a male display as a sign of a worthy mate does the same. There are of course other factors involved, which open these closed loops to the outside world. In the case of the peacock and peahen I have specified one of these in the differential predation on peacocks and conceptualized this as the enchaining of one feedback loop (predator and prey) with another (male and female birds). But to think of the selective history whereby peacocks developed elaborate tail feathers and peahens came to admire them as a closed system is not only analytically productive but also an accurate historical description as far as it goes.
Outside any biological cycle, however, there will always be other elements that determine or control it but are not in any significant way altered by it. They form the conditions for a feedback loop but are unaltered by the feedback. An obvious example for most biological systems is the climate. Sudden fluctuations in Upper Pleistocene global temperatures changed (p.33) human ecosystems but were effectively untouched by humans’ changed behaviors in response to them. The climate changes stood outside the alterations brought about by all the feedback cycles of which humans formed a part. Other obvious examples of such external elements are astronomical cycles, tectonic shifts, and volcanism. Three million years ago, the joining of the North and South American continents by the volcanic emergence of the Isthmus of Panama resulted in the so-called Great American Interchange of countless species. It altered all their cyclic relations with their ecosystems and drove many of them to extinction—notably the dominant predators of South America, the flightless “terror” birds. But the volcanism itself was unchanged by those alterations, not part of the feedback cycles it altered.
In control systems theory, elements sending a signal into a system from outside it are called feedforward elements. These external elements have the capacity to determine and alter in radical ways the systems they interact with. Defining these elements, however, will in many instances not be an absolute, black-and-white affair but a question of relative scales of systems and controls, and this is especially true in biological situations. The global climate, for example, might at first thought seem always to be a feedforward element. In considering local, small ecosystems this is for all practical purposes true; the impact of the ecosystem of a pond on the world’s climatic system is insignificant. But the example today of anthropogenic climate change is a clear instance of a single species remaking the global climate, in effect ushering what had been a feedforward element into the workings of its own feedback cycles (Tomlinson 2017); and this is not the first instance of such biotic remaking of the global climate. We can imagine scales between the pond ecosystem and human climate disruption where it will be difficult to judge whether an element plays a feedforward role, untouched by local feedback dynamics, or enters into those dynamics.
An additional complexity of feedforward elements will play a central role in the model of human evolution I offer later. In certain circumstances, feedback cycles can generate elements that come to stand outside them with emergent, organized dynamics of their own. These can then function as if they were controlling, feedforward elements, altering and determining the systems from which they arose with little change to themselves. In the broad view, these can hardly be thought of as pure instances of feedforward, since they took shape historically out of the very feedback dynamics they now seem to control; perhaps “pseudo-feedforward” or “quasi-feedforward” is a more accurate designation. But at a later historical moment, when their autonomous organization has taken full form, their operation can effectively mimic the impact of a true feedforward control. (p.34) We will see that hominin cultures came at a certain point to generate an abundance of such feedforward controls; these I call epicycles.
My definition here of the concept of feedforward is the one sanctioned in classic control theory, but it is a definition often misunderstood. In discussions of biological cycles and evolution in particular, the phrase “feed forward” is habitually used as a synonym for positive feedback, the “forward” thus distinguishing it from negative feedback, and it is even sometimes used to connote a simple causal chain, A—› B —› C, where there is no feedback at all. The distinction of feedforward from positive feedback or causal chains without feedback is, however, a categorical one, and the confusion is important to avoid.