eduKateSG · How Science Works
Choose your reading route
Follow both directions of reciprocal selection, calculate an original interaction model and test coevolution against simpler alternatives.
Distinguish interactions from reciprocal selection · Chapters 1–3 · Start this route
Calculate changing payoffs and read temporal evidence · Chapters 4–6 · Start this route
Include community context and test both causal arrows · Chapters 7–10 · Start this route
Full contents · Check your understanding · How Science Works master guide
A plant’s defences can influence which herbivores leave descendants. Changes in those herbivores can then alter which plant variants reproduce successfully. Each population can become part of the other’s changing selective environment. When evolutionary responses are reciprocal, the process is called coevolution.
The word demands more evidence than simply showing that two species interact. Pollination, predation, competition or parasitism can occur without a demonstrated evolutionary response in both partners. One species can adapt to another while the second changes little, or both can respond independently to the same climate.
A strong explanation therefore follows two arrows: how variation in partner A changes selection on B, and how variation in B changes selection on A. It then asks whether heritable change actually followed those pressures across generations.
This lesson extends Behavioural Ecology within the How Science Works master guide. Its calculations are original paper examples. Published host–parasite studies are discussed only as evidence, without culture, infection, evolutionary-manipulation or treatment procedures.
Contents
Distinguish interactions from reciprocal selection · Chapters 1–3
1. Interaction, selection and evolution are different claims
Calculate changing payoffs and read temporal evidence · Chapters 4–6
4. Partner frequencies change which variant is favoured
5. Evolutionary escalation is only one possible outcome
6. Comparing partners from different times can reveal hidden change
Include community context and test both causal arrows · Chapters 7–10
7. Communities can change the direction of reciprocal selection
8. Spatial and physical environments alter the possibilities
CHAPTER 1 OF 10 · Distinguish interactions from reciprocal selection
1. Interaction, selection and evolution are different claims
Start with an ecological interaction: an animal feeds on a plant. That observation establishes contact and resource use. It does not yet establish that the animal imposes different reproductive outcomes on heritable plant variants.
The next claim concerns selection. Suppose plants with one defence trait leave more surviving offspring in the presence of the animal than plants with another trait. The difference suggests selection associated with the interaction, provided relevant alternatives and confounding variables have been addressed.
Evolution adds another requirement. The relevant trait variation must have a heritable basis, and the population’s inherited composition or trait distribution must change. A plant temporarily increasing a defence in response to feeding can be a plastic response without an evolutionary change across generations.
Reciprocity adds the second population. Do plant differences also alter which herbivore variants reproduce, and does the herbivore population evolve in response? If only the plant side has been demonstrated, the evidence supports an evolutionary response to an interaction rather than the complete reciprocal process.
A useful diagram contains four boxes: A’s heritable variation, B’s reproductive differences, B’s heritable change and A’s altered reproductive differences. Add a further arrow toward A’s evolutionary response. Label separately which links are measured and which remain hypotheses.
The diagram is deliberately more demanding than two arrows between species names. It forces the explanation to identify traits, reproductive consequences and change through time instead of using “coevolution” as a decorative synonym for living together.
CHAPTER 2 OF 10 · Distinguish interactions from reciprocal selection
2. Why a well-matched pair is not proof by itself
A flower and its visitor may appear closely matched in shape. A predator and prey may have corresponding performance traits. Such patterns are reasonable starting points for a coevolutionary hypothesis, but several histories can produce a match.
One partner could have adapted to a relatively unchanged feature of the other. The two might have encountered one another only after their traits evolved independently. Environmental filtering might bring together organisms already suited to the same conditions. Plastic responses can also create a match without a genetic response in both populations.
Shared history introduces another alternative. Related organisms can resemble one another because of common ancestry, and related interacting groups may share habitats. A correlation among their traits should not automatically be treated as many independent instances of reciprocal adaptation.
The appropriate question is comparative: what would we expect if only one partner evolved, if neither evolved or if both changed because of a third factor? Evidence becomes stronger when the study distinguishes these alternatives rather than merely describing a convincing-looking pair.
A photograph records present compatibility. A reconstruction of coevolution requires information about variation, selection, inheritance and history. These are not mutually exclusive approaches; the photograph can guide the question, while experiments and comparative data test the proposed explanation.
CHAPTER 3 OF 10 · Distinguish interactions from reciprocal selection
3. Use two fitness records for one interaction
An original hypothetical example involves two plant variants, P0 and P1, and two herbivore variants, H0 and H1. The variants are assumed heritable for the purposes of the model. The chosen numbers are illustrative expected reproductive outputs, not measurements from a species.
First record plant performance as expected surviving offspring per plant. When paired with H0, P0 produces 80 and P1 produces 110. When paired with H1, P0 produces 60 and P1 produces 90. In the absence of herbivores, P0 produces 120 and P1 produces 100.
The plant comparison shows a trade-off. P1 has a twenty-offspring disadvantage without herbivores, but a thirty-offspring advantage over P0 with either herbivore variant in this example. A trait can therefore be favoured in one interaction context while carrying a cost in another.
Now make a separate record for herbivore reproduction. On P0, H0 produces 40 expected offspring and H1 produces 35. On P1, H0 produces 10 and H1 produces 30. H1 is five offspring behind H0 on P0 but twenty ahead on P1.
Do not subtract plant offspring from herbivore offspring as though the two species share a single numerical fitness currency. Each comparison is within one population under defined conditions. The paired records describe how the other species changes the relative performance of its variants.
These numbers establish reciprocal selection in the hypothetical model. They do not themselves show observed evolution, and they do not imply that the best immediate feeding performance necessarily predicts lifetime reproductive success in nature.
CHAPTER 4 OF 10 · Calculate changing payoffs and read temporal evidence
4. Partner frequencies change which variant is favoured
Let p be the frequency of P1 among plants encountered by the herbivores. Assume encounters sample plant types in proportion to those frequencies, with no preference, spatial clustering or density effect. The expected reproductive output of H0 is 40(1 − p) + 10p, or 40 − 30p.
For H1 it is 35(1 − p) + 30p, or 35 − 5p. Subtracting the two gives an H1 advantage of −5 + 25p. H1 is favoured when p exceeds 0.2, the variants tie at p = 0.2, and H0 is favoured below that threshold.
The number 0.2 is derived from the invented payoff values. It is not a biological threshold to apply to a plant–herbivore system. Its purpose is to show how one population’s composition changes the selective environment experienced by another.
At p = 0.5, the herbivore outputs are 25 for H0 and 32.5 for H1. If both herbivore variants initially have frequency one-half and reproduce in proportion to these outputs, the next H1 frequency is 32.5/(25 + 32.5), approximately 0.5652.
For a simultaneous plant calculation, let half the herbivores be H1. P0 then averages 70 offspring and P1 averages 100. Starting with equal plant frequencies gives a next P1 frequency of 100/(70 + 100), approximately 0.5882.
These are one-step deterministic frequency updates with equal starting frequencies and simple inheritance. They omit mutation, drift, migration, age structure and density dependence. They demonstrate a possible coupled response; they do not establish that the full system will cycle, escalate indefinitely or reach a particular long-term equilibrium.
CHAPTER 5 OF 10 · Calculate changing payoffs and read temporal evidence
5. Evolutionary escalation is only one possible outcome
An arms-race picture emphasises successive improvements in defence and counter-defence. It can be useful when later variants overcome earlier partners while retaining broad performance against them. But coevolution does not require endless increases in a single trait.
Trade-offs can make specialisation more important than universal improvement. A variant that performs well against one partner type may perform poorly against another. As partner frequencies change, the identity of the favoured variant can change too.
Negative frequency-dependent selection offers one route to continuing turnover. A host type that becomes common may favour opponents particularly effective against it. Once that opponent spreads, rarer host types can gain a relative advantage. The resulting dynamics need not be a ladder with each step permanently above the last.
Even stable average performance can conceal change. If both sides keep evolving, their contemporaneous interaction outcome may remain similar while their performance against partners from other times differs. This is one meaning of the Red Queen idea in coevolutionary research.
However, oscillating population numbers are not by themselves evidence of coevolution. Predator and prey abundances can cycle through ecological processes without heritable trait change. Likewise, a mathematical model that cycles demonstrates a consequence of its assumptions, not proof that a natural pair follows those assumptions.
The scientific task is to identify the measured trait dynamics and interaction outcomes. The label “arms race” should follow that evidence rather than substitute for it.
CHAPTER 6 OF 10 · Calculate changing payoffs and read temporal evidence
6. Comparing partners from different times can reveal hidden change
A time-shift comparison asks how a population from one period interacts with partners representing earlier, contemporary or later periods. Conceptually, this creates a matrix rather than a single present-day measurement.
For a paper diagram, place host sampling times along the rows and opponent sampling times along the columns. Each cell records a defined interaction outcome. The diagonal compares contemporaries; off-diagonal entries reveal how outcomes change when the partners come from different times.
If later opponents perform better against earlier hosts, that can support a directional change under a suitable comparison. If performance peaks against particular contemporary or recent partner types and declines against others, a matching or fluctuating process may fit better. No one matrix pattern uniquely identifies every underlying mechanism.
The 2007 study of Daphnia and parasites archived in pond sediment used the persistence of dormant stages to investigate historical interaction dynamics. Comparisons across time supported rapid parasite adaptation to host populations and were examined alongside a model of frequency-dependent coevolution.
The value of that evidence is temporal resolution. The study could ask about interactions among historical combinations rather than reconstructing change from one contemporary snapshot. It does not make every host–parasite system an identical Red Queen process.
Environmental carry-over, differences among archived samples and the choice of performance measure still require attention. For broader distinctions among infectivity, virulence, resistance and tolerance, use the existing Parasitology guide. Those quantities should not be collapsed into one undefined score of “winning”.
CHAPTER 7 OF 10 · Include community context and test both causal arrows
7. Communities can change the direction of reciprocal selection
An interacting pair lives within a community. A third species can change resource availability, predation pressure or which traits most strongly affect reproduction. Coevolution can therefore vary across locations even when the focal pair occurs in each place.
A 2001 study of crossbills and lodgepole pine examined this possibility by comparing settings with and without red squirrels. The researchers connected cone traits, bill traits and feeding performance with different selective environments. They interpreted the patterns as evidence for geographically variable reciprocal adaptation.
The important lesson is that a competitor can change which interaction dominates selection. A trait that matters strongly to crossbill feeding in one setting may be shaped mainly by a different seed predator elsewhere. The focal species pair is not isolated from the rest of the food web.
Such geographically variable interactions are often discussed as a mosaic, with locations differing in reciprocal selection and evolutionary history. Migration can connect these locations, moving variants among them and complicating any expectation of perfect local matching.
A map should therefore show measured locations and relevant community differences, not colour every region as either permanently coevolving or permanently inactive. The strength and direction of selection can change over time.
This also limits simplistic claims that two species must share a single universal evolutionary relationship. A species name can encompass populations experiencing different partners, climates and historical starting conditions.
CHAPTER 8 OF 10 · Include community context and test both causal arrows
8. Spatial and physical environments alter the possibilities
A 2022 bacterial–phage study documented diversification into multiple interaction types, including broader ranges and trade-offs associated with changed partner use. The result supports a richer picture than one uniformly stronger host type followed by one uniformly stronger opponent.
The authors also stated important design limits. Their comparison with well-mixed systems relied on earlier studies rather than a perfectly matched control, and spatial structure, movement, population size and evolutionary duration could not all be treated as interchangeable. Those qualifications prevent a strong claim that one environmental feature alone caused every difference.
Physical conditions can change reciprocal selection too. A 2025 plant–butterfly experiment measured traits, interactions and fitness in ambient and hotter environments. In that studied system, evidence of reciprocal selection found under ambient conditions was not detected in the hotter treatment.
This result concerns selection under the tested conditions. It is not a direct demonstration of many generations of reciprocal evolutionary change, and it does not establish that warming weakens every coevolutionary interaction. The distinction between a selective pressure and an evolutionary response remains essential.
Together, these studies show why the environment belongs inside the explanation. It can alter encounter patterns, costs, benefits and available variation. Coevolution is not a self-contained contest operating independently of temperature, space and community structure.
CHAPTER 9 OF 10 · Include community context and test both causal arrows
9. Build an evidence test for both arrows
Consider a fictional observation: over several years, plant defences increase and herbivore feeding structures change. A coevolutionary interpretation is plausible, but the two trends alone do not establish reciprocity. Both could be responding to drought, a new competitor or a change in which populations were sampled.
A stronger analysis separates four questions. Is variation in each trait heritable? Does each partner’s trait change the other’s relative reproductive success? Did inherited traits or allele frequencies change over generations? Do comparisons exclude important shared drivers and unilateral explanations?
A common-environment comparison can help distinguish inherited differences from immediate environmental responses. It is not automatically sufficient: maternal effects, developmental history and other carry-over effects may persist. The interpretation must match what the design actually standardises.
Comparing a changing partner with a fixed reference can also help. If one population changes while facing an unchanged partner, that demonstrates a different causal situation from two populations continually altering one another’s selective environment. Neither should be hidden under the same broad description.
For a paper worksheet, use the hypothetical payoff records from Chapters 3 and 4. Identify which cells show selection on the plant, which show selection on the herbivore and which reveal a cost. Then remove the herbivore-fitness record. Explain exactly why the remaining plant data cannot establish the reciprocal claim.
Finally, distinguish independent replicate populations from repeated measurements within one population. More observations improve resolution, but they do not manufacture independent evolutionary histories. A precise estimate of one history can still leave alternative causes unresolved.
CHAPTER 10 OF 10 · Include community context and test both causal arrows
10. Check reciprocity before telling an evolutionary story
Is coexistence enough to establish coevolution? No. The explanation requires reciprocal evolutionary responses, supported by evidence about selection, heritable variation and change in both partners.
Why can a defence be costly but still spread? Its benefit in the relevant interaction can outweigh its cost under those conditions. In the original model, P1 performs worse without herbivores but better in their presence.
Why does H1 become favoured above a plant frequency of 0.2? The invented payoff comparison gives an advantage of −5 + 25p. Its sign changes at p = 0.2. The threshold belongs only to that teaching model.
Does similar present-day performance mean no evolution occurred? Not necessarily. Both partners can change while their contemporaneous outcome remains similar. Time-shift comparisons can reveal information hidden by the present-day average.
Must coevolution produce mutual benefit or endless improvement? No. Antagonism, trade-offs, specialisation and changing partner frequencies can generate different outcomes. Evolution has no conscious goal or promise of universal progress.
The durable habit is to follow both arrows and preserve the distinction between interaction, selection and evolutionary response. Continue with Behavioural Ecology or return to the How Science Works master for related explanations of variation and ecological change.
Contents · Previous chapter · Continue to Behavioural Ecology