TLDR? Just read the bold bits!

I started with Darwin's Origin of Species, and then read Dawkins's first three books The Selfish Gene, The Extended Phenotype, and The Blind watchmaker (I just started River out of Eden).

Dawkins talks of Gradualists and Punctuationalists as if their views are incompatible but my initial understanding of Darwin was that both of these are true. My understanding is that if you look at an individual line of heredity then you will see a very consistent and gradual change through the generations (at the mutation rate), but if you look at the population as a whole you will see "periods of rapid evolution".

To me this seems easy to reconcile. Imagine a gene for fur thickness in polar bears, if each nth generation has a 'random' mutation for thicker or thinner then for each line of heredity you'll see a random walk about the starting point. Most of them will go up a bit, then down a bit, and end up close to where they started, some will be a bit further away, and a small number will be far away (much thicker or thinner fur) in a Gaussian distribution. So any random sample from the population that happens to become a fossil is most likely to be in the middle (i.e. unchanged over many generations).

Now if there's a sudden environmental change, or a one-off freak event such as a very cold winter, say that only the 10% with the thickest fur survived this cold winter (and perhaps a very small number of others who got lucky). Now very quickly within a few generations, that thick fur gene will dominate and as the population grows back to whatever size the environment can sustain it will be present in much greater absolute numbers than before the selection event. When looking at random samples of the population throughout this period, you might conclude that there has been a "period of rapid evolution". If the change is more gradual than a single sudden event, for example it gets colder over 500 years, each year the ones with the thinnest fur die off and the thicker fur genes get more popular in the population. This would appear to be a very rapid period of evolution, but really all the evolving happened gradually, long before the selection event.

My question really is what's wrong with this view of things? Dawkins seemed to anticipate all my thoughts and objections while I read his books and provided explanations, except for this one. This kept nagging at me the whole time and he never mentions it at all. I can't imagine I'm the first person to think this way (especially since Darwin himself pretty much said this) so why doesn't Dawkins explain it (either as correct or as incorrect with the reasons why)? Can someone please explain it to me either way?

  • $\begingroup$ Wlcome To Bio SE! It may help to bold-out the important parts, such as a specific question. It's a quite wordy question. Good luck! $\endgroup$
    – AliceD
    Jan 23, 2015 at 3:31

1 Answer 1


If I understand your question correctly, you are essentially distinguishing between the timescale on which the variation in the character of interest is generated, and the timescale on which selection prunes the population based upon that variation. The argument is then that "all the evolving happened gradually, long before the selection event" because that is when the variation was generated.

To me, this seems to be a misuse of the term "evolution", which is usually defined as a change in trait or allele frequencies in a population over time. Under that definition, much of the evolution does occur via the selection event. Before, you have a wide range of trait values; after, only a narrow selected set of trait values. In such a model, selection indeed is the major cause of allele frequency or trait value change.

In practice, both processes are usually acting in concert; variation is being produced by mutation and recombination and eliminated by selection at comparable rates so that the net amount of variation in the population is at an approximate steady state. This is essentially the model Darwin espouses, though note that Darwin does not have our modern understanding of heredity and therefore is not able to explain how variation is generated or preserved.

Additional note: When thinking about models of the kind that you describe, it is also key to understand that if you have a large number of loci affecting your trait (e.g. coat thickness), then the Gaussian distribution prior to strong selection will not include the extreme phenotypic values. If there are 20 diploid loci interacting additively to determine coat thickness and the mean individual has 20 thickness alleles and 20 thinness alleles, no individuals in a reasonably sized population will initially have all 40 thickness alleles, or even 39 or 38 etc. Selection for thickness will change the mean, and will also result in individuals outside of the initial range of variation. If the mean is 30, you may well get an individual with 38 or 39 or even all 40 thickness alleles. So once selection occurs, new more extreme phenotypes will suddenly appear. Nothing strange has happened in terms of rates of generating new variation; in fact this is just a consequence of reassorting the current alleles under the new allele frequencies.

  • $\begingroup$ So why are they presented as incompatible and opposing factions? Was this previously the case but now it's widely recognised that they work together? Did I just misunderstand? $\endgroup$
    – jhabbott
    Jan 23, 2015 at 9:17
  • $\begingroup$ Regarding the multiple loci - doesn't this just narrow the distribution? Meaning that the same arguments hold true, even more so in fact. $\endgroup$
    – jhabbott
    Jan 23, 2015 at 9:18
  • $\begingroup$ Generation of variation and selection on variation involve a very different scale (microevolution) than do graduated versus punctuated evolution (macroevolution). In the microevolutionary context, generation of variation and selection on variation have always been seen as mutually compatible, and indeed both necessary for evolution by natural selection to proceed. $\endgroup$
    – Corvus
    Jan 23, 2015 at 16:58
  • $\begingroup$ With regard to multiple loci, the idea is that multiple loci have a greater effect than any one locus could. If a single locus model and a multilocus model both spanned the same range of phenotypes, sure, you'd expect a narrower distribution for the multilocus case. But when each locus has an additive effect delta, we obviously have a broader ranger of effects and a larger variance in the multilocus case. $\endgroup$
    – Corvus
    Jan 23, 2015 at 17:01

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.