Population bottleneck models accuracy with known minimum bottlenecks?

Wondering if anyone is aware of any studies that may have looked at confirming the accuracy of population bottleneck models by determining the minimum bottlenecks for a known species with known bottleneck populations and times?

Ideally, a third party would collect all the sample DNA data and anonymize it somehow – perhaps focusing only on one gene and/or a small number of genes where the species cannot be readily identified – and then send to several different groups using different models to see what minimum population size and date they come up with.

An example species might be something like bison in North America (plains and/or wood), where the population was once 20 -30 million, then went all the way down to under 1,000 in the late 1800’s, then shot all the way back up to 500,000 in only about 100 years at present. Yes, humans helped revive them (after almost killing them off), but if we’re talking about using this study to look at the potential accuracy of these models for human bottlenecks, seems like there would be some similar intervention dynamics within human populations as well. And yes, cattle were admixed with the bison populations, but perhaps this is analogous to Neanderthal admixing with humans? :slight_smile: I believe wood bison in Canada may have less cattle admixing (if any), so maybe this sub-species would be better.

Since this bottleneck is such a short time ago and the diversity is so low, wouldn’t this be a slam dunk for population models?

Even better, is there some known very small bottleneck, say of some invasive species introduced by humans into an isolated island – where hundreds of years ago humans brought in a known and very small population…maybe even a population of two??

@swamidass: Did some searches and could not find anything similar being done, but let me know if you are aware of this or your thoughts.

However, if the group(s) doing the models already knew in advance the species and population minimums – skeptics may question whether the models were just tweaked to fit the known bottleneck(s).

I see that they determined the bottleneck for Florida Panthers via the genetic analysis, but don’t see if they can confirm this with any record-keeping data. It just seems like the genetics models are all that they have this this case:

Any other species that would be a good test? A small population introduced by humans hundreds of years ago on some remote island would seem ideal?

Again, to me, seems like this should be a slam dunk verification for population models, if done well. However,most of these would be so recent and so intertwined with human impacts that maybe they would not have much applicability to the accuracy of human bottlenecks that are 50k to ~100k years ago.

1 Like

Fazal Rana has written an article on this subject in which he points to studies which seem to falsify some population models. I am giving the main quote and link to his article below-

In my book Who Was Adam? I discuss three separate studies (involving mouflon sheep, Przewalski’s horses, and gray whales) in which the initial populations were known. When the researchers measured the genetic diversity generations after the initial populations were established, the genetic diversity was much greater than expected—again, based on the models relating genetic diversity and population size.2 In other words, this method failed validation in each of these cases. If researchers used the genetic variability to estimate original population sizes, the sizes would have measured larger than they actually were.


1 Like

I’m not sure why Rana cites this paper in his article:

Rana says of this study:

Of specific interest is a study published in 2012 by researchers from Finland. These scientists monitored the genetic diversity (focusing on 14 locations in the genome consisting of microsatellite DNA) of a population of white-tailed deer that were introduced into Finland from North America in 1934.5 The initial population consisted of three females and one male, and since then has grown to between 40,000 to 50,000 individuals. This population has remained isolated from all other deer populations since its introduction.
Though the researchers found that the genetic diversity of this population was lower than for a comparable population in Oklahoma (reflecting the genetic bottleneck that occurred when the original members of the population were relocated), it was still surprisingly high. Because of this unexpectedly high genetic diversity, size estimates for the initial population would be much greater than four individuals. To put it another way, this population size method fails validation—one more time.

In the article, the Finnish authors say (my emphasis):

Observed H was in line with predictions from an individual-based model where the genealogy of the males and females in the population were tracked and the population’s demography was included.

Indeed, their Figure 1 shows their model encompassing the observed population size history of the population.

His citation of this study is also misleading:


Rana says of this study:

Recently, conservation biologists have identified another factor influencing genetic diversity that confounds the straightforward application of the equations used to calculate initial population size: long generation times. That is, animals with long generation times display greater-than-anticipated genetic diversity, even when the population begins with a limited number of individuals.6
This finding is significant when it comes to discussions about Adam and Eve’s historicity. Human beings have long generation times—longer than white-tailed deer. From a creation model perspective, these long generation times help to explain why humanity displays such relatively large genetic diversity, even though we come from a primordial pair. And it suggests that the initial population size estimates for modern humans are likely exaggerated.

The authors say:

All sampled individuals were mature adults, thus representing progeny resulting from reproductive events of the former generation that occurred 20–30 years ago. Despite the fact that the population has suffered an important demographic decline in recent years, old individuals that were analysed in this study apparently still represent most of the genetic diversity (both in terms of heterozygosity and occurrence of rare alleles) that characterized the species prior to demographic reduction.


As explained above, the contemporary Ne estimate given by the linkage disequilibrium method is not statistically representative of the current generation, but the parental generation that created our sample, that is adult individuals that were living 20 – 30 years ago when the census size was likely substantially higher than today. The census size estimated in 2000 represents the actual number of adults in the population. Assuming that the population of copper redhorse has been through a reduction in population in size during the preceding years, the Ne estimate from the linkage disequi- librium method would be higher than what would be found in the current generation, thus creating a dispro- portionate ratio. Overall, our results exemplifies that the absolute Ne/N ratio should be used very carefully for conservation purposes, and particularly so in species with long generation time.


Under the assumption that the population size was dramatically reduced in the past 30 years, as surveys since the species’ recognition in 1942 have indicated, it could be too recent an event to detect the genetic consequences (Comité de rétablissement du Chevalier Cuivré 2004).

The authors’ point is that in the case of these fish, they have a generation time that is close to or exceeds the period of population decline from their previous population size to the size measured by census today. As a result, of course sampling adults today and doing genetic studies won’t give us an accurate measurement of the population today! If the population really is declining, as surveys apparently suggest, it’s no surprise that the genetics of adults born prior to that decline will give us a larger population size than we see today.

It’s a similar story for the white-tailed eagle study he cites:

The authors say:

A bottleneck lasting about 20–30 years is equivalent to about two white-tailed eagle generations. Simulation of a bottleneck comprising 30 reproductive pairs (which corresponds to the situation in several European populations during the 1970s; see §1) shows that only about 4% of the original heterozygosity is expected to be lost in an eagle population during that time (figure 2).

I really don’t see how Rana think these 2 studies could be relevant to a scenario where a primordial pair of humans were created many thousands of years ago. He seems to be under the impression that the papers say “longer generation times = inflated estimates of ancient population sizes”, but they say nothing of the sort.


I have not read the papers (except for skimming through and understanding very little)… so I can’t really comment on it.
I guess Dr Rana will have to give his reasons for the conclusion.

What kind of accuracy would you give to such calculations… such as those which say the minimum bottleneck possible based on current genetic diversity would be at least such and such no: so many years ago…

Are you aware of any validation studies across any kind of meaningful time scales (can it be be validated across hundreds of thousands of years?)

1 Like

I think Dr Rana was referring to this comment/finding-

General population genetic theory predicts that a sudden reduction in effective population size leads to an exponential decay in heterozygosity with a loss rate determined by the effective population size [6], [7], [8]. When ignoring mutations,
where the heterozygosity at time t ( Ht ) is given by the initial heterozygosity H 0 of the founders declining with a rate inversely related to the time-specific effective population size Nei [7],[8]. It has been recognised that this theorem can be oversimplistic when considering natural populations. For example, many populations have retained more of their diversity after a bottleneck than was predicted [9]. In particular, the predictions of classic theory (eq. 1) do not hold in organisms with overlapping generations (iteroparous organisms), because the loss of alleles is then much reduced [10]. In contrast to the classic theory, an individual-based population genetic model predicts that final heterozygosity is largely independent of the initial heterozygosity H0in the founding population; heterozygosity may well increase after founding

Basic idea being that an initial population with a lower number and hence less genetic variation can still produce a final population with more genetic variation.( Than expected…)

1 Like

Yes, than expected by “classic theory”, but reference #10 that the authors cite shows that we’ve known that there are confounding factors like overlapping generations since at least 1977. Unless current methods used to estimate ancient human population sizes haven’t progressed in the last 40 years, this isn’t a valid objection to modern analyses.


Thanks for this, I’ll check out these papers, but as I mentioned, not sure what would prevent the authors from tweaking their model parameters to match the historical numbers. Would be so much better if the data was somehow made anonymous and sent to several different teams, like sometimes is done with new radiometric dating samples. Don’t tell them the answer, see if they all come to the same conclusion first.

I can’t quantify the accuracy of a method that will applied in different way to different data in different studies. i think in general they are quite accurate, as efforts are made to account for confounding factors. Of course they’re not perfect, and a particular dataset can be consistent with several historical scenarios, so researchers will always have to weigh up all the available evidence before they arrive at the final estimate. This final estimate will obviously also have associated error bars of different confidence intervals. Regardless of our ability to determine ancient population sizes though, there is enough evidence to show that we didn’t come from a single population of 2 specially created individuals as Rana and RTB believe in though. The closest you could get is to an ancient bottleneck to 2 individuals from an ancestral population.

1 Like

Yes, I think we can agree here. This might be Rana’s main point. I found the below passage interesting. Highlighting what intrigued me-

However, the difference in heterozygosity was minimal (0.742 in Oklahoma and 0.692 in Finland; implying 93% was retained), although this reduction was marginally significant. In terms of classic population genetic theory (eq. 1), this high level of retention of heterozygosity in a population presumably founded by four individuals is unexpected. Nevertheless, our individual-based simulation of population establishment where the genealogy of the males and females in the population are tracked allowed calculation of a predicted heterozygosity which closely matches the observed heterozygosity in the current Finnish white-tailed deer population. Because the introduced white-tailed deer population enjoyed rapid initial population growth, it spent a relatively short time period at a small size, which has allowed the retention of relatively much genetic diversity.

Even starting with 4 individuals, the final population retained 93% of the genetic diversity found in the main population… and the rate of population growth/ how short the bottleneck was in duration impacts the final genetic diversity.

I agree.
Can you point to any literature on what the improvements are… and how they have been validated?
Also it’s interesting to note that these calculations all assume selection didn’t have any significant impact on the genetic variation.

Yes, this is a known effect in very short bottlenecks - not a lot of heterozygosity is lost. See this paper from 1986: https://onlinelibrary.wiley.com/doi/pdf/10.1002/zoo.1430050212

One improvement is that we don’t use microsatellites any more, we use polymorphism data from whole-genome sequences.

If you want an overview, here’s the overview of a special issue in the journal Heredity with links to several papers discussing recent advancements:

1 Like


If I am reading it correctly, the final population retained 93% of the genetic diversity found in the 4 individuals that founded the population. Of course, those 4 initial founders have a small percentage of the genetic diversity of the larger deer population.

Heterozygosity is also not the same as genetic variation. If you founded a human population with 2 people you would have, at most, 4 HLA-A alleles. Even if these 4 alleles were evenly mixed in the subsequent population, that is still just 4 alleles. There are over 4,500 HLA-A alleles in the modern human population. A recent bottleneck for the human population definitely doesn’t work.


It’s 93% compared to another population in Oklahoma which didn’t go through a bottleneck.
They calculate the heterozygosity as 0.742 and 0.692 respectively.