Natural Selection in Ecological Context

Near Cambridge, in the 2000s, birds took dark peppered moths more often than pale ones: each day a dark moth survived at about 0.91 times the rate of a pale one. That single number took six years of releases to estimate. This lesson asks what it takes to make a claim like it, and what such a number does and does not tell you.

The argument: variation, inheritance, differential reproduction, environment

Darwin and Wallace’s papers were read together at the Linnean Society of London on 1 July 1858. The argument they made still has the same parts, and a claim that selection is acting needs all four.

1. Variation in a trait. Individuals differ in beak depth, tail length, colour or gall size.
2. Inheritance. Some of that variation passes from parents to offspring.
3. A measured difference in fitness. Fitness is reproductive success relative to other variants in a given environment. You rarely measure lifetime reproduction directly, so you measure a fitness component: survival over a stated period, mating success, or fecundity (number of eggs or young).
4. A named agent in a stated environment. A predator, a parasitoid, a competitor, a choosy female, a drought. The same trait can be favoured in one place or year and disfavoured in another.

Think of selection as a sieve. The sieve does not make the grains; it sorts what is poured in. A drought does not create deep beaks. It removes more of the birds whose beaks are shallow, among the variants already present.

Keep two steps apart. Selection is a difference in fitness among phenotypes within one generation. Evolutionary response is the change across generations, and it happens only to the extent that the selected variation is inherited. A trait can be strongly selected and not change at all if its variation is environmental. How large the response is, and how allele frequencies move under a given fitness difference, belong to population genetics and complex traits. Here the job is to show that the selection is real and to name its cause.

Selection also need not keep pointing one way. On Daphne Major in the Galápagos, Peter and Rosemary Grant measured selection on the beaks and bodies of two finch species from 1972 to 2001, and it ranged from consistently one-directional to oscillating (testing evolutionary explanations). There was no single permanently favoured beak.

Directional and stabilizing: gall flies caught between wasps and birds

Plot relative fitness against a trait and the shape of the curve names the mode. Directional selection favours one end of the trait distribution. Stabilizing selection favours intermediates over both extremes.

The goldenrod gall fly, Eurosta solidaginis, lays an egg in a bud of tall goldenrod, and the plant grows a round gall around the larva. In fields in central Pennsylvania, Abrahamson, Sattler, McCrea and Weis (1989) followed what killed larvae in galls of different sizes. Small galls left larvae vulnerable to parasitoid wasps that lay their eggs in them. Large galls were more often opened by birds. Larvae in intermediate galls had the highest survival: stabilizing selection, with two named agents pressing from opposite sides.

The same data also held a directional component. The mean directional selection intensity favoured larger galls by 0.21 phenotypic standard deviations, a standardized shift in the trait, not a death rate. So one trait, in one set of fields, carried both components at once. The modes are descriptions of a fitness curve, not boxes a trait must fit into.

Disruptive and frequency-dependent: crowded sticklebacks and rare guppies

Disruptive selection favours both extremes over intermediates. Theory predicts it when individuals compete for the commonest resource and those that use something different do better. Bolnick (2004) tested this with threespine sticklebacks, Gasterosteus aculeatus, in paired enclosures in each of five lakes, one at low and one at high density. Because he lacked lifetime reproduction, he used body size and relative gonad mass as proxies for potential fecundity. Disruptive selection inferred from those proxies was consistently stronger at high density. Manipulating density made crowding a tested cause, within the limits of the proxies.

Reading the curve takes care. Martin and Pfennig (2012), working on Mexican spadefoot tadpoles, Spea multiplicata, in 15 ponds in Arizona and New Mexico, applied the established caution that a significant positive quadratic term alone does not show a fitness minimum among the trait values actually sampled. If the low point lies outside the data, the curve is directional selection that accelerates. Look at the fitted curve, not only the coefficient.

Negative frequency-dependent selection favours a phenotype when it is rare. To test it you must separate rarity from the phenotype itself, because a colour could do well for reasons that have nothing to do with how common it is. Olendorf and colleagues (2006) did this with male guppies, Poecilia reticulata, in three natural populations in Trinidad streams. They set male colour-pattern types at about 3 common to 1 rare, and reversed which morph was rare between replicate pools. After 15 or 17 days, depending on the site, rare males had higher estimated survival, whichever pattern was rare. Recapture stands in for survival only if missing fish died rather than moved, and predators forming a search image for the common type remained a possible explanation, not a demonstrated one. The next generation was not measured.

Sexual selection: the widowbird tail

Sexual selection is selection through mating success, by competition for mates or by mate choice. Male long-tailed widowbirds, Euplectes progne, carry very long breeding tails. Comparing males that happened to differ in tail length could not separate the tail from age or territory quality, so Andersson (1982) changed the tails of males in Kenya. Males given longer tails attracted more females to nest on their territories than males with normal or shortened tails, and males with shortened tails held their territories as long as the others. The tail affected mating, not territory holding, which points to female choice. What the experiment does not show is why females prefer long tails, or the net lifetime fitness of a long-tailed male.

Peppered moths revisited: what was measured

In the peppered moth, Biston betularia, the dark carbonaria form became common around British industrial cities in the 19th century and declined after air pollution fell in the late 20th. The colour difference is inherited, and van’t Hof and colleagues (2016) traced carbonaria to a transposable-element insertion in the gene cortex. Older predation experiments had been criticised for how moths were released. Michael Majerus repeated the work near Coton, Cambridgeshire: 4,864 moths over six years, checked four hours after sunrise, with moths that had gone presumed eaten. Cook, Grant, Saccheri and Mallet (2012) analysed his data and estimated selection against carbonaria at s = 0.091 per day (likelihood 95% interval 0.028 to 0.157).

relative daily survival of carbonaria = 1 - s = 0.909

Worked problem. What does 0.909 per day do to the make-up of a resting population? The starting numbers here are invented for the exercise. Suppose 1,000 typica and 1,000 carbonaria moths, typica daily survival 0.80, and no migration, reproduction or change in rates.

1. Carbonaria daily survival. 0.80 x 0.909 = 0.7272.
2. Expected survivors after one day. Typica: 1,000 x 0.80 = 800. Carbonaria: 1,000 x 0.7272 = 727.2.
3. Carbonaria share among survivors. 727.2 / (800 + 727.2) = 727.2 / 1,527.2 = 0.476, or 47.6%, down from 50%.
4. Check by cancellation. The 0.80 and the 1,000s cancel: 0.909 / (1 + 0.909) = 0.909 / 1.909 = 0.476.
5. Five days. The ratio compounds: 0.909^5 = 0.621, so the share is 0.621 / (1 + 0.621) = 0.621 / 1.621 = 0.383, about 38%.

carbonaria share after t days = 0.909^t / (1 + 0.909^t) (equal starting numbers)

Three cautions. A fractional survivor such as 727.2 is an expectation, not a moth. The 0.091 is relative and per day, at one site over those years: not an annual decline, not an absolute mortality, and not a figure for all of Britain. And a change in morph share among survivors is not yet a change in allele frequency; that needs the dominance and inheritance model in population genetics. Majerus’s moths also “disappeared”: only about 26% of disappearances were witnessed bird attacks.

Selection against carbonaria, s per day 0.091 Day 1 Resample

	daily survival	expected survivors	random draw
typica	0.8000	800.0	–
carbonaria	0.7272	727.2	–
carbonaria share	47.6%		–

In this sketch: 1,000 typica and 1,000 carbonaria to start (all invented numbers), typica daily survival 0.80, no migration, no reproduction, rates constant. Fractional survivors are expectations. s is relative and per day. Resample: in this sketch each moth survives independently with the probability shown (a sampling model of this sketch, not from the study).

Rule to keep: carbonaria share after t days = (1 − s)^t / (1 + (1 − s)^t) for equal starting numbers. A change in morph share among survivors is not yet a change in allele frequency; that is left to the Genetics lesson.

Design problem

Students propose that birds take rare colour morphs of a prey less often than common ones. Design a field test.

One way to do it

Use several comparable pools or plots. Make morph A rare in some replicates and common in others, for example at 1 A to 3 B in half and 3 A to 1 B in the rest, so that each morph is rare somewhere. Keep total density, habitat and season the same across replicates, so that only frequency differs.

Mark every individual, release, and recapture after a set interval, such as two weeks. State the assumption that an individual not recaptured has died rather than emigrated, and reduce emigration with barriers or by searching neighbouring habitat.

The prediction is that whichever morph is rare survives better, in both sets of replicates. If one colour does better everywhere, the result is directional selection on colour, not frequency dependence. To test a mechanism, watch predators and record whether they take the common morph more than its share, the pattern a search image would produce.

Finally, keep the conclusion to its scope. One season of survival in these sites shows frequency-dependent selection on survival. It does not show an evolutionary response; for that you would need evidence that the colour is inherited and data on the next generation.