In the book written by John Endler Natural Selection in the Wild p. 4 Natural Selection in the Wild p. 4 it says that even if condition a, b and c are met, evolution by natural selection might occur,

[...] , but not necessarily, [...]

See last paragraph of the image. Endler, J. A. 1986. Natural Selection in the Wild. Princeton University Press.

I’m wondering in which case it’s possible that everything is in place for natural selection to occur, but is not having any effect.

It seems to me that if every condition of natural selection are met, there will absolutely a change in the phenotypic distribution? Why does the text say evolution is not guaranteed to occur?

  • $\begingroup$ if all the living things are killed for instance there is no natural selection. if the environment does not affect the outcome of reproduction, i suppose that natural selection is not occurring, but its hard to imagine such a situation $\endgroup$
    – shigeta
    Dec 30, 2015 at 7:32
  • $\begingroup$ Hi @mbeausoleil - could you respond to the comments under remi.b's answer - we could do with clarity $\endgroup$
    – rg255
    Jan 20, 2016 at 22:13

2 Answers 2


Evolution, a change in the trait distribution of a population, occurs by four mechanisms; drift, mutation, migration, and selection. E.g. These mechanisms cause a change in the mean of trait, or variance in a trait etc.. Just to summarise the conditions you gave:

a) Trait $i$ has variation

b) Trait $i$ covaries with fitness ($\omega$)

c) Trait $i$ is heritable from parent to offspring

Genetic drift is the random loss of genetic variance from the population by stochastic processes. Mutation (generally) generates new genetic variation within populations. Migration can lead to loss (emigration) or gain (immigration) of genetic variation from a population, as variants come or go from a population.

Selection is mechanism underlying adaptation. This is what is in condition b of the conditions you gave. We can predict the response to selection using some simple quantitative genetics in the form of the (multivariate) breeders equation.

$\Delta z = G\beta$

Where $\Delta z$ is the response in the trait, $G$ is the genetic (co)variance matrix, and $\beta$ is the selection.

In a simple toy example we could look at a single trait, $i$. We allow $i$ to satisfy all of the above conditions. This means that both $G$ and $\beta$ are non-zero, and therefore, $\Delta z_i$ will also be non-zero. Just putting some random numbers in to the equation to make it clear, if $G = 0.5$ and $\beta = 0.8$ then;

$\Delta z_i = 0.5 \times 0.8 = 0.4$

Mutation, drift, and migration could all oppose or distort the effect of selection. In other words, they can cause the actual and predicted response to be different from one another (i.e. $\Delta z_i \neq 0.4$), and there will not necessarily be a change in the phenotypic distribution.

However, selection is also factor that could cause a difference between the predicted and actual response to selection in trait $i$. This is because genetic correlations can arise among traits, either by linkage (close proximity in the DNA) or pleiotropy (when one gene affects more than one trait). For example, we may see no response at all in $i$ ($\Delta z_i = 0$) when both $G_i$ and $\beta_i$ are non-zero, if selection on a second unmeasured trait, $j$, opposes selection on $i$ ($\beta_i = -\beta_j$), and trait $j$ has equal variance ($G_i = G_j$) and is perfectly covarying with $i$, $cov_{i,j} 1$. Approximately this would be;

$\Delta z_i = (G_i \times \beta_i) + (G_{i,j} \times \beta_j$)

$\Delta z_i = (0.5 \times 0.8) + (0.5 \times -0.8) = 0$

This is a major issue in the study of evolutionary biology. Genetic correlations can cause severe disparity between the actual and predicted response to selection and univariate methods are insufficient as a result. More studies adopt multivariate methods these days, though these are generally limited to just a handful of traits so its still not perfect. In reality methodological and logistical constraints present a huge obstacle to being able to predict $\Delta z_i$.


What does the sentence below mean?

As a result of this process, but not necessarily, the trait distribution may change in a predictable way.

To my understanding, but not necessarily means that other processes than natural selection can affect the trait distribution in a predictable way.

Other than natural selection, what affects the trait distribution in a predictable way?

I am not trying to make an exhaustive list but I am just providing here two obvious examples.

  • An intense gene flow (migration) from a nearby population will also cause the trait distribution in a local population to change in a predictable way.

  • Genetic drift, will reduce genetic variance and therefore reduce the variance of the trait distribution. While the change of the mean of the trait distribution is not predictable under genetic drift, its change in variance is predictable.

Note that the book has been published in 1986 and the vocabulary and science is probably slightly outdated. If you are learning about evolutionary biology, you can probably find better source of knowledge. For example Understanding Evolution is a free online introduction to evolutionary biology. There are many textbooks that offer very good introductions to evolutionary biology.

  • 2
    $\begingroup$ I'd add genetic correlations - selection on one trait (a) may oppose the selection on another genetically correlated trait (b), so when studying a all conditions would be met but no change (or lesser than expected, or opposite to expectation, change) occurs $\endgroup$
    – rg255
    Dec 27, 2015 at 21:54
  • $\begingroup$ Sure I could add that. However, as you said, this is NOT an example of when something else than natural selection can affect the trait distribution in a predictable way. It is an example of when selection is counteracted by something else. $\endgroup$
    – Remi.b
    Dec 29, 2015 at 17:31
  • $\begingroup$ Upon reading the question again, I think the user is asking why evolution doesn't always occur (in the predicted way) if the conditions for natural selection are satisfied. A question of why doesn't selection always give a response, rather than how else can predictable evolutionary change occur. $\endgroup$
    – rg255
    Dec 30, 2015 at 7:04
  • 1
    $\begingroup$ Hum... indeed. However, there is contradiction in the sense of the question between the actual question and the citation that the question is supposed to tackle. As a consequence the post would deserve being closed as unclear. $\endgroup$
    – Remi.b
    Dec 31, 2015 at 22:49
  • 1
    $\begingroup$ Agreed, need the OP to come back! $\endgroup$
    – rg255
    Jan 1, 2016 at 0:58

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .