Jump to ContentJump to Main Navigation
Darwinian AgricultureHow Understanding Evolution Can Improve Agriculture$

R. Ford Denison

Print publication date: 2012

Print ISBN-13: 9780691139500

Published to Princeton Scholarship Online: October 2017

DOI: 10.23943/princeton/9780691139500.001.0001

Show Summary Details
Page of

PRINTED FROM PRINCETON SCHOLARSHIP ONLINE (www.princeton.universitypressscholarship.com). (c) Copyright Princeton University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in HSO for personal use (for details see http://www.universitypressscholarship.com/page/privacy-policy). Subscriber: null; date: 15 August 2018

Learning from Plants, Ants, and Ecosystems

Learning from Plants, Ants, and Ecosystems

(p.177) 11 Learning from Plants, Ants, and Ecosystems
Darwinian Agriculture

R. Ford Denison

Princeton University Press

Abstract and Keywords

This chapter examines the interactions between trees and fungus-growing-leaf-cutter ants, arguing that we can learn much from natural communities as long as we don't mistakenly assume perfection. It shows that natural selection can improve the fitness of each participant in multispecies interactions, regardless of the impact on the community as a whole. The chapter begins with an evolutionary perspective on chemicals as either signals, cues, or manipulation. This is followed by a discussion of the fungus-growing ants, with particular emphasis on their natural strategies for pest control, and the use of biotechnology for biological control of pests. The chapter concludes by stressing the importance of natural ecosystems in providing essential context for understanding the sophisticated adaptations of wild species, before applying them to agriculture.

Keywords:   trees, fungus-growing ants, natural selection, multispecies interactions, chemicals, pest control, biotechnology, leaf-cutter ants, natural ecosystems, agriculture

We can learn much from natural communities, if we don’t mistakenly assume perfection. Natural selection tends to improve the fitness of each participant in multispecies interactions, regardless of the impact on the community as a whole. Understanding this, can we learn how to optimize such interactions in an agricultural context?

Signals, Cues, and Manipulation

The previous chapter explained how gaseous chemicals that benefit aphids in one context (warning them of predators) can harm them in others (scaring them away host plants). I will start this chapter with an evolutionary perspective on chemicals as either signals, cues, or manipulation. Then I will revisit the fungus-growing ants introduced in chapter 1, focusing on their natural pest-control methods. Last, I will emphasize the importance of natural ecosystems in providing essential context for understanding the sophisticated adaptations of wild species, before applying them to agriculture.

As discussed in the previous chapter, some wasps use aphid alarm signals to find aphids to parasitize. Other wasps respond to volatile chemical “distress signals” released by plants being attacked by caterpillars.335 Are these really signals—that is, are they really beneficial to both sender and receiver? If so, how did they evolve?

The term signal is sometimes used loosely. An extreme example can be found in recent use of the term to refer to any chemical released by one microbe that has nonlethal effects on another microbe, even when the same chemical is lethal at higher concentrations. For example, antibiotics made by some bacteria kill other bacteria at high concentrations. At low concentrations, however, antibiotics may induce bacteria to aggregate into layers known as biofilms.336

Biofilm formation is often seen as a cooperative activity. But my former PhD student Will Ratliff and I have suggested337 that biofilms formed in response to antibiotics may be analogous to selfish herds,338 where animals try to push into the center of a crowd to escape predators. Is a cell (p.178) joining a biofilm in response to an antibiotic really doing so to benefit the other cells already there, or to escape from the antibiotic, whatever the consequences for its new neighbors?

It helps to use different terms for information-bearing molecules that have different effects on the fitness of senders and receivers. After all, those fitness effects will determine how production of these molecules, and responses to them, will change under natural selection. The term signal should be reserved for messages (including chemicals) that benefit both sender and receiver. A chemical that merely happens to provide useful information to others, but isn’t produced for that reason, is better described as a cue.339

Which term we use can depend on whose fitness we are considering. Aphids may signal to their relatives, but those signals can be used as cues by parasitic wasps. Natural selection will enhance the ability to detect and respond to useful information, so it is not surprising that wasps can detect aphids some distance away. In this case, producing the volatile chemical has conflicting effects on the fitness of the aphids, helping their kin avoid predators, but also increasing the chance of detection by parasitic wasps.

An individual that benefits by producing a chemical message that harms the recipient, typically by providing misleading information, is engaging in manipulation. The bolas spider, which makes moth pheromones to lure male moths close enough to catch, is an example.

Natural selection among recipients will tend to decrease their susceptibility to harmful manipulation, perhaps by reducing their responses to certain chemicals. Selection in the species doing the manipulation, meanwhile, will tend to increase the strength of the signal produced, if doing so increases fitness of the sender. This is another example of an evolutionary arms race. In contrast to manipulation, selection will increase responses to useful cues, responses that have also been called eavesdropping,17 and to mutually beneficial signals. Chemicals released by legume roots that attract symbiotic rhizobia are a textbook example of a signal.340 By attracting rhizobia and housing them in its root nodules, plants gain access to atmospheric nitrogen. A single rhizobial cell founding a nodule can produces millions or billions of descendants inside, many of which may eventually escape into the soil, so responding to plant signals can greatly increase the fitness of a rhizobial cell. Sender and receiver can both benefit, justifying the use of the term signal.

Responding to plant recruitment signals may not always benefit a rhizobial cell, however. What if there are more predatory protozoa around roots, and many more aspiring nodule-founders than there are nodulation opportunities?341 Do legumes, like the farm-owners in The Grapes of Wrath, advertise more openings than actually exist, to attract a surplus (p.179) of applicants? Do rhizobia manipulate each other, overproducing the quorum-sensing (see glossary) signals they use to estimate their own population density, thereby encouraging competitors to disperse to less-crowded areas?262 The differences between cues, manipulation, and signals will be important as we consider their possible use in improving agriculture.

Plant Defense by “Bodyguards”

When they are attacked by caterpillars, some plants release volatile chemicals that attract caterpillar-parasitizing wasps.335 Recruiting wasp “bodyguards” benefits both the plants, which get protection, and the wasps, which get live hosts in which to lay their eggs. With these mutual benefits, these chemicals qualify as signals.

But how did this interaction evolve? It probably started with wasps eavesdropping. Among parasitic wasps, only those that find hosts will reproduce. Therefore, each successive generation of wasps will tend to be better at detecting cues that lead them to their hosts. Aphid alarm signals are an example of such cues.

Another example comes from wasps that parasitize butterfly eggs. Some male butterflies include a chemical anti-aphrodisiac with their sperm, decreasing the chance that a female will mate again, which would dilute the first male’s contribution to the next generation’s butterfly gene pool. Some parasitic wasps have evolved the ability to detect this anti-aphrodisiac chemical. The wasp uses this cue to find a pregnant butterfly, then follows her—the tiny wasp may even hitch a ride on the larger butterfly—until she lays her eggs, and then parasitizes them.342

Given the propensity of parasitic wasps to use all available cues, we can guess how plant signaling to wasp bodyguards may have evolved. Plants wounded by caterpillars would have released some volatile chemical cues just through leakage, even before any natural selection for producing signals. Wasps with behavioral mutations that made them tend to seek out these chemicals were more likely to find caterpillar hosts and reproduce, spreading the alleles for responding to those chemicals.

Once large numbers of wasps began using wounded-plant volatiles as cues, mutant plants that produced more of these volatile chemicals would get more protection from caterpillars, relative to plants that produced less of the same chemical. So a chemical that started as a cue used by wasps became a signal from plants to bodyguards. Some beetles that eat those plants, however, still use the same chemical as a cue.343

Some plants have turned signals into manipulation. Certain orchids “cry wolf,” releasing volatile chemicals similar to those produced by (p.180) wounded plants, even when they are not being attacked. These chemicals attract parasitic wasps. During their futile search for caterpillars to parasitize, they pollinate the orchids.344

There are probably many more examples of chemical information exchange among plants and beneficial and harmful insects not yet discovered. Can we use what we learn about signals, cues, and manipulation to reduce pest damage in agriculture?

Biotechnology and Biological Control of Pests

As noted at the end of the previous chapter, advances in biotechnology have made it possible to change the chemical cues released by crop plants, in ways that might enhance pest control. For example, we may be able to increase the ability of crops to attract beneficial predatory or parasitic insects. Would this approach provide more-lasting protection against pests, relative to the current approach of adding insecticidal toxins to crops?

Biological control of pests by predators or parasites that survive and reproduce on a farm is potentially “evolution-proof.” Pests evolve ways to hide from their predators or parasites, but that imposes selection for better search strategies by the predators or parasites, because they depend on their prey or hosts to reproduce. The pests may take the lead in this arms-race for a year or two,230 but coevolution by the predators will eventually catch up.

This coevolution is often missing, however, when we apply beneficial insects that don’t survive long on the farm, or when we use crops that make their own insecticides, just as it is when we spray insecticides. For example, crops genetically engineered to produce the Bt toxin impose selection for Bt-resistant pests, as discussed in the previous chapter. But Bt-resistant pests in a farmer’s field don’t effectively select for improved Bt genes in the crop plants. This is because most farmers purchase seed every year, grown in distant fields where those particular evolved pests are absent. Similarly, beneficial insects that are purchased and applied to control pests on a farm won’t coevolve with those pests unless they manage to survive and reproduce on that farm.

If farmers plant seed produced on their own farms, there will be some ongoing selection for pest resistance in their crops. There may also be selection for undesirable plant traits, however, including less within-crop cooperation, as discussed in chapter 8. Letting natural selection work on your farm might maintain pest resistance (and otherwise improve adaptation to local conditions), but it could also lead to wasteful competitive traits like excess height.

(p.181) What if instead of developing crops that make defensive toxins, we develop crops that send gaseous signals that attract beneficial predators or parasites of crop pests?333 It would probably be easiest to develop crops that produce such signals all the time, rather than only when those pests are present. An individual plant sending such signals, surrounded by nonsignaling plants, would presumably suffer less pest damage. This could also be true for a small group of plants. If you have a few broccoli plants in a community garden that send out bodyguard-recruiting signals, maybe you could attract beneficial predators and parasites to your plants from your neighbor’s plot.

But what happens when all the plants on a farm produce the same signal, even in the absence of pests? Beneficial predators and parasites would then have no useful information to guide them to their prey or hosts. This simpleminded approach to using natural predators and parasites would therefore undermine, rather than strengthen, biological control. Rather than using biotechnology to make crops cry wolf, therefore, I would focus on making sure crop genetic improvement programs don’t inadvertently eliminate the natural cues that already enhance biological control.

More sophisticated ways of combining biotechnology with biological control might work, however. If most plants in a field produced a volatile chemical that repelled pests (alarm signals used by those pests, perhaps), while a small patch of plants produced attractive chemical lures, maybe most of the pests could be drawn to the smaller patch. With this push-pull strategy,334 predators or parasites might be attracted to the smaller patch by the abundance of prey or hosts, reducing their need to search the whole field.

Alternatively, biological control insects raised elsewhere could be released directly into these small attractive patches. Remember, however, that predators or parasites raised elsewhere will not have coevolved with potential prey or hosts on a particular farm. If natural selection continuously improves the ability of pests to escape introduced biological-control insects, while purchased biological-control insects stay the same every year, it’s pretty clear who will eventually win that evolutionary arms race.

Pest Control in Ant Agriculture

A related problem may limit the ability of fungus-growing ants to control harmful fungi in their nests. The best-known of these pest fungi is Escovopsis, which isn’t eaten by ants but attacks their fungal crop. As mentioned in chapter 1, ants apparently use “pesticides” to control Escovopsis. (p.182) Cameron Currie and colleagues at the University of Wisconsin showed that these pesticides are produced by bacteria that live on the bodies of the ants themselves.23

If the pest fungi are killed by a chemical, but the chemical is made by bacteria, is this chemical control (analogous to pesticides) or biological control? The bacteria apparently don’t interact directly with Escovopsis, so it seems more like chemical control.

Whatever we call it, there are important evolutionary questions here. If Escovopsis evolves resistance to the bacterial toxins, will the bacteria coevolve to produce new toxins that overcome that resistance? In other words, are the bacteria more like coevolving beneficial predatory wasps established on a farm, or are they more like purchased wasps that don’t coevolve with their target pests? To answer this question, we first need to ask why do the bacteria make these antifungal chemicals.

Is there another tragedy of the commons problem here? Making and excreting antifungal chemicals must have some metabolic cost to an individual bacterial cell. Therefore, if a random mutation in one bacterial cell knocks out antifungal production, that cell should reproduce somewhat faster than nonmutants around it. Within a few days (many bacterial generations), bacterial cheaters that don’t make the antifungal should displace those that do. This should happen, at least, unless bacteria producing the antifungal somehow benefit preferentially, relative to those that don’t.

If the bacteria abandon the production of antifungal compounds, that could eventually let Escovopsis destroy the ants’ fungal crop, causing the ants to starve. That might be bad news for the bacteria. But as noted in chapter 3, natural selection depends on current conditions, not future consequences, and on individual benefits, not collective ones. Could making antifungals somehow provide a short-term benefit to individual bacteria, outweighing its metabolic cost?

If the bacteria ate Escovopsis, there could be strong selection among bacteria to produce the antifungals needed to kill their fungal “prey.” This would be similar to biological control, where a predatory insect population evolves to counter evolving defenses in its prey. But these bacteria appear to eat ant excretions, not Escovopsis. The chemicals the bacteria produce therefore seem more like nonevolving chemicals (or nonevolving beneficial insects) made or raised in a lab and applied in the field. If lab-raised predators don’t evolve to counter prey evolution in the field, what would impose selection for antifungal production in bacteria living on ants?

Recently, a possible explanation for the evolutionary persistence of this antibiotic production was published by Currie’s group, although they didn’t interpret their results the way I do.345 They found that the ants’ bodies can host yeasts (single-cell fungi) as well as bacteria. These (p.183) yeasts reduce the growth of the bacteria. If yeasts attack or compete with the bacteria, then there would be an immediate, individual advantage to producing antifungal compounds that suppress yeast. Antifungals that suppress yeast might also suppress Escovopsis, as a useful side effect.

In the short run, the yeast reduce growth of bacteria, limiting their production of antifungals that suppress Escovopsis. That was the conclusion emphasized by Currie’s group. Over a longer period, however, selection imposed by yeast could maintain antifungal production by the bacteria, keeping them from losing the ability to suppress Escovopsis.

But if selection for antifungal production is imposed by a fungus other than Escovopsis (by a yeast, for example), why doesn’t that antifungal also harm the fungal crop? Or does it? Ulrich Mueller (formerly Currie’s major professor) and colleagues recently reported that bacterial “secretions kill or strongly suppress ant-cultivated fungi,”346 not just Escovopsis. This is what we would expect, if there is no direct selection for activity specifically against Escovopsis. The ants can only hope that coevolution of the bacteria with the yeast will maintain effectiveness against Escovopsis as well.

Or do ants use other methods? The answer appears to be “it depends.” There are many species of fungus-growing ants, which differ in various ways. A group of researchers collected nine of these ant species from a national park in Panama and compared them.347 Patches of antifungal-producing bacteria were seen (with a low-power microscope) on some ant species, but not on others. The family tree for these ant species has been worked out, and the researchers were able to map the presence or absence of these bacteria onto that phylogeny, as in the transfer RNA example in chapter 3. The last common ancestor shared by all of these fungus-growing ant species apparently hosted the bacteria, but that trait was lost twice, in the ancestors of five of the nine species.

Even bacteria-less ant species aren’t helpless against Escovopsis, however. Those ants apparently make antibiotics themselves, in their own special glands. They then spread the antibiotics on their gardens and on each other. These antibiotics kill Escovopsis, but they can also harm the fungal crop. Therefore, behavioral adaptations (using these chemicals only when necessary) may be as important as the biochemical ability to produce them. This need for caution in applying toxic chemicals applies to human farmers as well. For example, herbicides applied to kill weeds in row crops sometime drift into orchards, harming valuable trees.

The mystery of how natural selection among bacteria that live on ants maintains production of Escovopsis-specific antifungals may have been solved: their antifungals are not so specific after all. This may also be true for antifungals made by the ants themselves, but those are used by the ants in ways that carefully target harmful fungi.

(p.184) Like an individual bacterium, an individual ant may not benefit from the effort needed to kill Escovopsis. But unlike a bacterium, a worker ant can have no direct descendants. Her genes reach the next generation only through the reproductive success of her mother, the queen. And her mother’s survival and reproduction depends on keeping the colony free of Escovopsis. So kin selection is key here.

An interesting correlation seen in the study just described is that the ant species that rely more on their own antifungals tend to have larger colonies. Furthermore, those colonies are also more genetically diverse, because the queen mates with more than one partner.347 A colony with more genetic diversity might produce a wider range of antibiotics, reducing the chances that fungal pests would evolve resistance to all of the ants’ chemical defenses.

Although the queens of some social insects, including ants, have multiple partners today, ancestral state reconstruction suggests that they are descended from monogamous ancestors.348 With monogamy, Hamilton’s r for a worker’s sister can be greater than for her own offspring. (The greater depends on an unusual genetic system found in some insects, which also makes workers less related to their brothers.142) Therefore, kin selection can favor workers feeding the queen’s offspring rather than their own. If the queen had multiple partners, on the other hand, then Hamilton’s r decreases for sisters. So worker bees and ants might not have abandoned reproduction, if their mothers hadn’t been monogamous.349,350 Once workers lost the ability to reproduce, however, queens were free to take multiple partners, perhaps benefiting pest control by increasing the genetic diversity of the colony.

If monoculture and toxic chemicals (carefully applied?) have worked for ants for 50 million years,19 can we conclude that these practices are sustainable? Maybe, if our definition of sustainability in a human context is “humans not going extinct.” But how often do individual ant colonies collapse? I don’t know the answer to this question. And, even if monoculture has met the minimal test of persistence, would polyculture be even better?

Unlike whole ecosystems, ant colonies compete against other ant colonies for resources like leaves. Have there been colonies that practiced polyculture that consistently lost in these competitions? If so, can we conclude that monoculture is more sustainable than polyculture?

Maybe not. The main reason ants use monoculture seems to be the tendency of different fungal genotypes to attack each other when in physical contact. Maybe an ant colony would reduce the risk of catastrophic crop failure if it grew more than one genotype of fungus, but they can’t grow them together. (The phrase “separating the sheep from the goats” has nothing to do with a tendency of these species to fight each other, but you get the idea.)

(p.185) Even growing different fungal genotypes in separate chambers might not work, because droppings from ants eating one fungal genotype trigger adverse reactions in other genotypes.351 Humans, however, have more options for mixing crops, because our crop plants don’t usually have these sorts of hostile interactions.

Another way in which ants protect their crops deserves attention. They grow them underground, physically isolating them from most potential pests. Human farmers can sometimes benefit from a similar approach. Each year, it seems, more of my brother’s farm is covered with inexpensive plastic greenhouses. This is partly to keep crops warm in cold weather, but it also reduces some pest problems. For example, his raspberries grown under plastic don’t get splashed with mud containing mold spores. If you’re in Oregon when raspberries are in season, stop by a farmers’ market and taste the delicious results!

New and surprising results on ant fungus farms are being published every year. For example, it was recently reported that although much of the nitrogen ants need is imported in the protein of leaves they harvest, some comes from nitrogen-fixing bacteria in their own gardens.352 Do ants have any way to favor bacterial genotypes that fix more nitrogen, analogous to the sanctions imposed by legume plants on their root nodule bacteria?

Given how many new things we are learning about ant agriculture, it may be too early to try to adapt their pest-control strategies for use on our farms. An approach that has worked for ants for millions of years might not work for us, but it’s certainly worth studying. How and why does it work? What factors are key to its success? To what extent are the challenges faced by ants similar to or different from the challenges faced by human farmers?

To really understand the strengths and weaknesses of ants’ agricultural practices, we need to study them as complex communities (ants, crop fungi, Escovopsis, two or more species of beneficial bacteria, yeasts, and maybe other species to be discovered), ideally in the context where they evolved: natural ecosystems. It is only in that context that we can fully understand their complex adaptations and determine their relevance to human agriculture.

Ecological Context and the Value of Natural Ecosystems

Ecological context is likely to be equally important in understanding natural selection’s other innovative adaptations in wild species. Those adaptations are coded in DNA, but we can’t understand them just by studying their DNA.

(p.186) As discussed in previous chapters, our justified confidence in nature’s ancient wisdom is based mainly on repeated competitive testing by past natural selection. Natural selection operates mainly to improve genes, secondarily to improve individuals and families (like ant colonies), and only sporadically (or as a side-effect) to improve entire multispecies communities. So it might seem that we could preserve natural selection’s innovations just by saving seeds, or even just DNA.

But our goal is not just to preserve nature’s library, but to read it, understand it, and apply it. Often, the understanding we need can come only from studying living organisms in the environments where they evolved.

We might guess, from the DNA sequence of a gene and comparisons with similar genes, that the gene codes for an enzyme that attaches a methyl group to some molecule. But could we determine, just from studying the DNA, or even from studying the living plant in isolation, that the function of the molecule is to attract wasp bodyguards? Probably not, unless we already knew about those bodyguards and which chemicals attract them.

Similarly, if phosphorus-scavenging proteoid roots hadn’t already been discovered through field research on plants growing in low-phosphorus soils, how would we decipher the functions of the genes involved so that we could make intelligent decisions about whether to transfer those genes to other plants? DNA and RNA sequencing are powerful tools353,354—their costs have dropped enough that I’m starting to use them in my own research—but much of their value will be lost unless we study adaptations in the context where they evolved.

This context-dependence of natural selection’s innovative solutions is a rarely recognized reason why it is important to preserve natural ecosystems. Currently, the most widely accepted reason to preserve natural ecosystems is that they provide various ecosystem services, such as purification of air and water.57

One problem with the ecosystem services rationale is that their economic value can be undermined by changes in land use. For example, the value of pollination services provided by wild bees in Costa Rican forest fragments (totaling 147 hectares) to a nearby 1065-hectare coffee plantation was once estimated at $60,000 per year,355 providing a strong economic argument for preserving those forest fragments. But then coffee prices dropped and the farm switched to growing pineapple, which doesn’t depend on bee pollination.356 Did natural forests suddenly become much less valuable?

What about endangered species? Rare snail darter fish, once featured in lawsuits that delayed dam construction, don’t do much for the ozone layer. Maybe they eat disease-causing mosquitoes, but less-rare fish eat (p.187) many more. The rarer a species gets, the fewer ecosystem services it provides, undermining the ecosystem services argument for preservation. If we value nature mainly for the ecosystem services it provides, should we let rare species go extinct, and focus on the few species that are endangered despite being abundant?

Not if we see wild species as a valuable source of ideas, as I suggest. Then rare species can be just as valuable as common ones.

For example, researchers studying fish that live in cloudy water have discovered that their eyes process polarized light in ways that let them see farther through the murk, a technology that may be adapted to help people driving or flying through fog.357 Some sharks have a special microscopic skin texture that greatly reduces drag; this discovery has already been applied to boat hulls, reducing fuel use.358 Sharks also had counter-current heat exchangers millions of years before humans invented this energy-conserving device. As far as I know, the snail darter hasn’t yet inspired any inventions useful to humans, but that may be because they haven’t been studied in enough detail.

If wild species go extinct, we may lose valuable ideas encoded in their DNA. If we preserve their DNA, but destroy the ecosystems where they evolved, we may never be able to figure out all the clever things their genes did. Not all of the problems solved by those genes will be relevant to human needs in agriculture, medicine, or engineering. But one good idea can be worth more than a huge pile of lumber.

I should mention one reason why mimicking the overall organization of a natural ecosystem might sometimes make sense, even with limited understanding of how that organization affects ecosystem productivity, efficiency, or stability. I have argued that the adaptations of individual wild plants have been repeatedly tested and improved by natural selection, while the overall organization of natural ecosystems has not. However, most of natural selection’s testing of individual adaptations took place in natural ecosystems.

So if we find an individual adaptation that we don’t understand, but that works so well that we decide to copy it blindly (perhaps by crossing a crop species with one of its wild relatives), do we also need to copy some aspects of the ecosystem where it evolved, to provide an environment where those adaptations will be most successful? If we don’t really understand why a particular adaptation works so well, we might not be able to predict which features of the natural ecosystem the adaptation needs in order to keep working. However, I would be very interested in hearing from anyone who has a good example of such a case.

(p.188) Comparisons among Ecosystems

Even if the organization of natural ecosystems hasn’t been consistently improved by natural selection or any other natural process, we may learn much from studying them. We shouldn’t expect ecosystem-level properties, like the number of species or how they are arranged in time or space, to be optimal. But neither should we expect natural landscapes to be as badly designed as some human-modified landscapes.

In a 2003 paper, we suggested studying natural ecosystems the way an educator might study the educational systems of other countries.28 Which countries are having the best results? What, if anything, do they have in common? This is a very different approach from asserting that a particular country has the best schools and then copying its system uncritically. Similarly, once we agree on how we want our agricultural landscapes to function, we can look for natural (and agricultural) ecosystems that best meet our performance criteria and try to figure out why those exemplary ecosystems work so well.

It’s important to note that natural ecosystems that perform poorly by agricultural criteria may be just as valuable as sources of information on what doesn’t work as those that are highly productive, stable, and efficient. We might even want to study natural ecosystems that perform especially badly, such as those with particularly severe boom-and-bust cycles of predators and prey, as examples to avoid.

This informational role is particularly important for remnant ecosystems that now occupy little land area. Why do we care how much carbon dioxide a remnant prairie takes out of the atmosphere, on a per-acre basis? There aren’t enough acres of prairie left for that to matter to the global atmosphere. The real value of remnant natural ecosystems is as a source of ideas and inspiration, both practical (as emphasized in this book) and aesthetic.


In chapter 1, I introduced fungus-growing ants to show that certain practices that are widely considered unsustainable (monoculture, toxic pesticides, and eating high on the food chain) have persisted in natural ecosystems for millions of years. But mere persistence is a weak criterion, relative to competitive testing. This is our second core principle, from chapter 4. For example, the persistence of wild-rice monocultures in natural lakes doesn’t prove their intrinsic superiority to more-diverse plant communities—lakes don’t reproduce based on successful competition against other lakes.

(p.189) But ant colonies do compete against each other, with the winners spawning a disproportionate share of new colonies. So the farming methods we see in ant colonies today have been subject to the competitive testing of natural selection. Therefore, according to the first core principle from chapter 4, we might not expect many opportunities for simple, tradeoff-free improvements in ant agriculture. What can we conclude about monoculture, the use of toxins to control pests, and using animals (or fungi) to convert plants to human food?

Because different strains of the ants’ fungi attack each other, converting fungal monoculture to polyculture might not qualify as simple. Ant colonies with polycultures of mutually compatible fungal strains may never have existed to compete against ant colonies that practice monoculture. So direct assessments of the effects of crop diversity, as intercrops or in rotation (see chapter 7), may be more useful than anything we could learn about crop diversity from ants.

As for toxins, their widespread use by plants to defend themselves against herbivores adds to the conclusion from ants—toxins can work. And, in some form, they can keep working for millions of years. In contrast to the apparent sustainability of natural toxins, human-designed toxins are often quickly overcome by pest evolution. So we may have much to learn from ants and plants about the sustainable use of toxins. Keep in mind, however, that our criteria are broader than those of natural selection. We want pesticides that keep working for decades, but without causing cancer or harming wildlife.

Last, if ants eat high on the food chain, should we? Maybe sometimes. The third core principle from chapter 4 is that a diversity of approaches can be a useful form of bet-hedging. This idea is explored in the final chapter.