I was wondering if we could think about experiments to falsify natural selection . According to this wiki : we could invalidate natural selection

“...if it could be shown that selection or environmental pressures do not favor the reproductive success of better adapted individuals”.

So, to test this hypothesis we would have to:

  • Have different individuals that are defined as more or less adapted to a given environment
  • Put them in that environment
  • See if the better adapted individuals have a better reproductive success. (I will avoid the word fitness because one can consider it already contains the principle of natural selection)

But is it possible to define a priori that an individual is more or less adapted? Of course, we could suppose a priori that because of some or other trait (higher longevity, higher stress tolerance, or whatever), an individual carrying that trait will be more adapted. But if we find out this individual does not have a better reproductive success in that environment, we could conclude it was just not the right trait to be better adapted and it would not invalidate natural selection.

Isn’t it, on the contrary, the fact that some individuals have a better reproductive success which allows defining them as better adapted a posteriori? There would then be some kind of a vicious circle.

Ok, so one could say that we do not need to define a priori better or less adapted individuals. We could just put individuals in an environment and see if at least some have a better reproductive success. That would mean that they are better adapted –and so that natural selection occurs-. But suppose we find no significant differences in individuals reproductive success: that could indeed mean that natural selection does not exist. But, well, that could also just mean that there was no selection pressure in that environment.

So if we want to test natural selection existence, we need either to get individuals that are a 100% sure better adapted than others; or environmental conditions that 100% sure exert selection pressures on organisms ; and we need to know this a priori. But, it seems to me that considering an individual as “adapted” or an environment as “selective” can only be done a posteriori if we already accept the idea of natural selection.

That’s why, to me, it is difficult to think of an experiment that would falsify natural selection. Is natural selection falsifiable ?

  • $\begingroup$ Short comment... Why do you think we have multi resistent bacteria? Let me tell you why... we are selecting those few in a group with higher resistance everytime we expose the group to an antibiotic. $\endgroup$ Aug 10, 2017 at 19:24
  • 2
    $\begingroup$ Okay, sorry for that I'll edit my post. Just to be clear however : I never implied that anything sucked and I never doubted about natural selection's validity. This was just a serious question about its epistemological status. $\endgroup$
    – Kjian
    Aug 10, 2017 at 20:10

1 Answer 1


What is natural selection?

Natural Selection (NS) is an evolutionary process causing a directional change in both a phenotypic trait and the allele frequencies associated with particular values of this phenotypic trait whenever there is a trait that

  1. Has a non-zero narrow-sense heritability $h_N^2$
    • See this post for the definition of heritability
    • Note that if $h_N^2 ≠ 0$, then it necessarily implies that the trait is variable in the population.
  2. Correlate with fitness variance in the population

You may want to have a look at this post (maybe in particular Lewontin's recipe explained in this answer) or have a look at an intro course to evolution such as Understanding Evolution by UC Berkeley if this is unclear to you.

You may also want to have a look at the post How is “selection” best defined?

Note about the environment

Note that the genotype - fitness covariance does not necessarily need to be dependent or correlated with a specific environmental variable. Many selection pressures are indeed independent of the environment. If one specific genotype is associated to heart failure, then the environment may matter little. This genotype will always be associated with lower fitness.

In other words, you do not need environmental change, or environment specific adaptations in mind to test NS.

Would evidence that what is needed for natural selection to happen does not exist falsify NS?


I don't think that an evidence that mutations do not occur would falsify NS because while NS needs genetic variation (fuelled by mutations), the process per say does not refer to the creation of new genetic variance via mutations. Showing that mutation do not occur would not per say falsify NS (I think) but it would make the point that NS cannot happen. A bit like, the absence of energy would not falsify the theory of gravity but it would prevents any gravity to ever happen.

The theory of evolution as whole would be falsified if one were to show mutations never occur (or occur at a much lower or much higher rate than currently observed) though.


Similarly, if one could show that $h_N^2 = 0$ for all traits (nothing is heritable), then this would also prevent any NS to ever happen

Fitness variance

Similarly, if one could show that there is no fitness variance in population or even that fitness is always independent of the gentic variance (in other words, fitness itself is not heritable), then NS would never happen.

How to falsify natural selection?

You could show that for a phenotypic trait $h_N^2 > 0$ and show that genetic variance correlates with fitness variance (fitness can be measured as the number of offsprings) and yet repeatedly fail to see any change in the phenotypic trait values (or frequency of the loci explaining variance for this trait) over time. Such observations would falsify NS.

  • $\begingroup$ I think my difficulty comes from the fact that it is hard to think a situation in which a heritable trait which provides fitness benefits would not bring to a change in genotypic or phenotypic frequency associated to this trait over time. I wonder if theoretical models have shown this possible. $\endgroup$
    – Kjian
    Aug 10, 2017 at 21:27
  • $\begingroup$ Well, no, I don't think a theoretical model has shown the opposite possible. But it is not because a cause to consequence is too easy to grasp that it makes its formulation unfalsifiable. All our models have been developed following some logical cause-consequence relationships. Some are hard to get an intuition from. NS has the advantage to be a fairly simple mechanism, easy to understand even for a layman. $\endgroup$
    – Remi.b
    Aug 11, 2017 at 1:24
  • $\begingroup$ I also wonder about something you said: can environnent really be independent on the genotype-fitness covariance ? If you put a terrestrial mammal with a heart failure gene and one without it at the bottom of the sea, they'll both have null fitness. Put them both in a futuristic world where every disease is cured before birth, they will have the same positive fitness. $\endgroup$
    – Kjian
    Aug 11, 2017 at 6:23
  • $\begingroup$ The genotype-fitness covariance is measured on an existing population. You cannot observe the effect of extreme environments not explored by the species range. But I agree this can be misleading. Of course, genotype-fitness covariance would necessarily be dependent upon the environment but not necessarily upon the experienced environment. $\endgroup$
    – Remi.b
    Aug 11, 2017 at 7:36
  • $\begingroup$ @Remi.b This is a good definition, but where in this definition is an ability to distinguish between natural selection and genetic drift? $\endgroup$
    – sterid
    Nov 4, 2017 at 21:46

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