What Genetics Says About Adam and Eve

For many Christians, worries about evolution center on one question: did the human race start with Adam and Eve? Given the importance of the Fall and original sin in many versions of Christian theology, the concern is not surprising; unlike many issues raised by evolution, this one seems to impinge directly on a core doctrine. Is the Genesis story—of a single couple living in a garden, who eat the forbidden fruit and bring sin into the world—is that story true or not? Can’t science answer this question for us? Well, no, not really. It turns out that Christians mean very different things by “the question” and read the Genesis story in very different ways. What science can do, however, is put some strong constraints on plausible answers to the question.

In particular, one of the things science can do is estimate how large our ancestral population was. That is, taking the whole modern human population together, how many ancestors did we have at different times in the past? The branch of science that can make that kind of estimate is population genetics, the mathematical study of genetic variation within a species—in this case our species. That’s relevant to Genesis because in the creation story (at least as I learned it growing up in a conservative Baptist church), Adam and Eve were created not too many thousands of years ago and were the sole ancestors of all modern humans. Or in the language of population genetics, our ancestral population size was exactly two in the quite recent past (at least as population geneticists think about “recently”). That means we can frame a meaningful scientific question about Adam and Eve: Is there any genetic evidence that our ancestral population consisted of only two individuals at some point? If so, when? If not, to what extent does the evidence actually rule out that possibility?

To assess the answers that genetics offers to these questions, it helps to understand how DNA can tell us anything at all about history. Let’s start with the genome: one complete set of human DNA, which contains the information needed to form a functioning human. Some of the DNA carries that information, while other bits are important for structuring and reproducing the DNA, and still others seem not to do anything at all, but all of it is contained in the genome’s 3 billion-long string of chemical ‘bases’. The bases come in four varieties, usually represented by their initials A, C, G, and T, and it is their order that carries information. We each have two copies of a full genome, one inherited from our father and one from our mother. In turn, we pass on half of our DNA to each child—but first, we mix up chunks of our two copies, in a process called recombination, and pass on the resulting mosaic. What our children inherit from us is an almost perfect copy of what we inherited from our parents, suitably mixed. Given how many times DNA has to be copied between the birth of one generation and the birth of the next, and how often DNA experiences chemical damage, it is astonishing how few errors occur in making the copies: perhaps one error (i.e. one mutation) on average every 75 million bases. Most of these mutations have no effect on our health or other characteristics; they can be silently passed on to future generations, to be joined by a new batch of changes each generation.

This almost-but-not-quite-perfect fidelity is why DNA provides such a powerful tool for probing the past. Because it changes, it creates a record of all of the ancestors it has passed through; because it changes very slowly, that record lasts a long time, preserving information about events that occurred thousands of generations ago. To understand what this record can tell us: think about me as a parent, passing on copies of a single segment of DNA to both of my sons. Their copies started out identical to each other, because they are copies of a single piece of my DNA; in the language of genetics, they share a very recent common ancestor. Once they are in different people, though, the copies will have independent trajectories, each accumulating its own distinct mutations and becoming more and more different from the other every generation. (Assuming my kids ever have kids of their own, that is. Take your time, guys—I’m not in any hurry.) Because new differences are added every generation, they can serve as a crude clock that measures how long it has been since their common ancestor: the more differences, the longer it has been. This is the core idea behind all of our inferences about our ancestral population.

With that genetic clock in mind, what do we see when we look at human DNA? Start with the simplest thing we can do: take DNA from randomly selected people and count the number of differences between their copies. In this way, we are measuring human genetic diversity, something that is now easy to do, thanks to public databases with complete genome sequences for thousands of people from around the world. When we do that comparison, we find that two copies of human DNA typically differ by about 0.1%; that is, one base out of every thousand is different. Given the mutation rate above—one mutation every 75 million bases per generation— this means that two copies of human DNA have typically been accumulating differences for around a million years.

To a population geneticist, a measurement of diversity like this can immediately be translated into an estimate of the long-term size of the human population. That’s because larger populations retain more genetic variation than small ones. (Think of it like this: If you and all of your ancestors live in one small village, then everyone in the village will be some kind of cousin. Your DNA will share a recent common ancestor with everyone else’s, and there will be very few differences between your genomes. In other words, small population = low diversity. On the other hand, if you compare DNA from people all over the world, there will be many differences because they are very distantly related. Large population = high diversity.) For us to have a genetic diversity of 0.1%, given our mutation rate, we must come from a population with a long-term size of something like 15 or 20 thousand individuals—the kind of size we would expect for a largish mammal like humans.

Taken at face value, then, human genetic diversity suggests a simple conclusion: we descend from a population of many thousands, and our DNA has been around and collecting mutations for at least a million years. So does that settle the matter? No, not really. For starters, the size we’re estimating is a long-term average, not the size at one particular time (technically, it’s something geneticists call the “effective population size”). More importantly, this estimate assumes that all the genetic diversity we see is the result of accumulated mutations, an assumption that clearly need not apply if we are taking seriously the idea that Adam and Eve were real and were specially created. In that case, we would have to consider the possibility that they were created with large amounts of genetic diversity built in—that their copies of the genome could have been quite different from one another, and it is those differences, passed down to their descendants, that we see when we compare modern genomes. In other words, our observed diversity is consistent with the usual scientific picture, in which our ancestors evolved gradually as a large population, but by itself it does not rule out special creation. Most of the differences we see today between genomes could represent genetic variants that were created in humans at the beginning.

We do not have to stop there, however: we can test the hypothesis that our observed diversity was created. For starters, we can see whether the genetic variants we see look like they come from mutations or not. To do that, we can rely on the fact that different kinds of mutations occur at different rates. For example, a ‘C’ base is more likely to change into a ‘T’ base than into a ‘G’, so we should find more places in the genome where some people have C and some T than places with C and G. (Check out this piece for a full explanation of the different classes of mutations and the patterns we expect from them)

Conveniently, we also have an easy way to distinguish between variants that must have arisen by mutation and those that could have been originally created: their frequency in the population. If we all descend from Adam and Eve, then any variants that we inherit from them should be widely shared between individuals throughout the world; we can call these “high frequency” variants, for which both bases are present in many genomes. Rare variants, those for which one base is found only in a few people, would mostly be the result of mutations that have accumulated since Adam and Eve—let’s call these “low frequency” variants. If low frequency variants really come from mutation while high frequency ones were directly created, we might reasonably expect them to have different characteristics. In particular, we would only expect the mutations to show the distinctive pattern described in the earlier blog post. But that’s not what we see. Rather, the identical pattern is seen regardless of frequency. On the left is the pattern we see in the low frequency variants (in this case, variants found in about 1% of all genomes), which is exactly that expected for new mutations. But the pattern on the right, which comes from high frequency variants (found in about 50% of genomes) is the same. (I’ve taken the data for both plots from the 1000 Genomes Project website.)

In other words, yes, it is a theoretical possibility that quite a bit of human genetic variation was created in a primordial couple, but the genetic evidence does not look that way. Instead, it looks very much like all of our variants come from mutation. And as we noted earlier, if the variation represents accumulated mutation, that accumulation has been going on for a very long time, on the order of a million years.

Categorizing variants by how frequent they are in the population is more than just a trick to answer questions about Adam and Eve—it is a powerful tool used by geneticists to probe the history of populations. It provides much more information about that history than simply counting differences, as we did above. The key fact here is that genetic variants always start off rare, since they come into existence as a change in one copy of DNA, one copy among thousands or millions (currently billions) in the whole population. That frequency does not have to stay the same, though: it can drift up and down from generation to generation. This happens because each person who has the variant can pass it on to one offspring—leading to no frequency change in the next generation—or to two or three offspring, or to none at all. In this way, the frequency can decrease or increase as more or fewer people happen to be carrying it. Wait long enough, and the variant will end up either disappearing entirely or taking over; i.e., it will eventually reach either 0% or 100% frequency. When that happens, that site in the genome will no longer differ between people.

This process of “genetic drift” happens slowly. Exactly how slowly depends on the size of the population. For example, if there are only 50 individuals alive, a new mutation in one of them will create a variant with a frequency of 1%, since it will be present in 1 out of the 100 genomes in the population (50 individuals x 2 genome copies per individual). If it is passed on to 3 children, say, its frequency immediately jumps to 3%. By contrast, if the population has half a million people in it, the variant starts out as one copy among a million (0.0001% frequency) and it will take very many generations to get up even to 1%. One consequence is that variants that are at high frequency are old—typically very old—if the population is large. In addition, there are usually more rare variants than common ones, since only a fraction of new variants drift to higher frequencies rather than being lost. We can predict just how many more rare variants there should be, and what the entire distribution should look like, for different population histories—that is the sort of thing population geneticists do for a living. For example, this is a simulation of what the distribution should look like if humans have had a constant sized population of 16,000 for millions of years:

As you can see, we expect far more low frequency variants than high frequency ones. (For those who care, the expected functional form is , where f is the frequency of the lower frequency base at each site.)

All we have to do, then, is tabulate all the genetic variants in a large number of people and see how common each variant is. This we can again do using the thousands of publicly available genomes and the millions of variant sites that have been found in them. For example, if you look at position 87,144,623 on chromosome 6, you will find that there is a genetic variant there. At that site, 6% of genomes have a C as the base while 94% have a T, so we say that that variant has a frequency of 6%. Doing that for every variant in the genome allows us to build up a picture of the distribution of frequencies. Here is what we find for genomes from across Africa (again using the 1000 Genomes data), plotted along with the the simulated curve shown above:

The curves look broadly similar, and they match up very well for higher frequencies. As we’ve just noted, that means they agree well for older variants. For less common variants, though, we find a lot more in real genomes than in the simulation. Recalling that a larger population = more variants, we can conclude that the ancestral population size increased at some point. To a population geneticist, then, this distribution looks like it comes from a population with a long-term size of about 16,000, but which has expanded in size more recently. That is a conclusion that fits with everything else we know about our recent history as a species: most African populations, like most populations worldwide, have increased greatly in size in the last 10,000 years or so, thanks to the development of agriculture.

It is worth noting that we see quite a different picture if we look at people with non-African ancestry. For example, here I’ve added data collected from various European groups:

Once again, the two curves are identical for the most common (= oldest) variants, but when we look at rarer variants, the Europeans have fewer than we expect from a constant-sized population. By the logic we just used, this must mean that the European population actually got smaller at some point. Is that plausible? Actually, it makes a great deal of sense. Our current understanding is that modern humans first appeared in Africa, some two or three hundred thousand years ago, and they have lived there ever since. Only comparatively recently, starting perhaps 60 or 70 thousand years ago, did a small fraction of that population start migrating out of Africa to populate the rest of the world. The small size of that migrating population—a population “bottleneck” in genetics language—left its mark in the frequency distribution we see here.

Population geneticists do not just sit and look at plots like these, of course; they study the distributions in detail and subject them to mathematical analysis. Their broad conclusions, though, are the same as described here. Many local human populations have been through bottlenecks at some point in their history, but the population as a whole, and in particular many sub-Saharan African populations, look like they have had populations at least in the thousands for hundreds of thousands of years.

A straightforward interpretation of that data is that our ancestors were part of a population in the thousands as far back as we can see.

In other words, the answer to our original question is no: genetic data does not provide evidence that our ancestral population ever consisted of a single couple. A straightforward interpretation of that data is that our ancestors were part of a population in the thousands as far back as we can see. That still leaves our followup question, though: Does this data actually rule out a single pair of ancestors? And remember, we have to answer that question while allowing for arbitrary amounts of genetic diversity in the original pair. Unfortunately, this is not a question the scientific literature is going to provide much direct help with; scientists are simply not spending a lot of time thinking about Adam and Eve and whether they fit in with genetic data. Instead, we have to rely on informal efforts and on general population genetics principles.

Whether genetic data is even remotely compatible with a bottleneck of two individuals—let’s call them Adam and Eve—depends a great deal on exactly when they would have lived. A single ancestral pair who lived 10,000 years ago or less, as a common traditional approach to biblical dating would have it, raises very different problems than a pair who lived 500,000 years ago. To see why, let’s look at that frequency distribution again. What would it look like if we all descended from a single pair 10,000 years ago? Here is a simple simulation of that scenario (along with the African data again):

For this simulation, I allowed as much created variation in the original couple as I could—enough to account for all of the higher frequency variants. (Note that I’m now ignoring the issue I raised above, that high frequency variants look like they come from mutations.) That’s because original variants all start out at a high frequency: if there are only four genomes (which is what two people carry), then every variant has a minimum frequency of 25%. Subsequent drift can smear out those high frequencies, but they cannot produce the characteristic smoothly rising curve at low frequencies that we see in data—that has to come from mutations. And to produce that curve takes both a long time and a large population. A large population, because small populations cannot maintain the levels of diversity we see, and a long time because variant frequencies drift slowly in a large population. As the little blip on the left of the simulation shows, 10,000 years is not remotely long enough to accumulate that many mutations.

How much time would be needed for enough mutations to arrive and drift up to higher frequencies? The shortest time I’ve been able to come up with, working with my own simulations, is about half a million years—anything short of that produces too little low frequency variation. Similar results were found by Ann Gauger and Ola Hössjer, who attempted a more systematic analysis with their own simulations, searching for a scenario that included special creation of Adam and Eve and was also consistent with genetic data. (They published their results in BIO-Complexity, the pro-Intelligent Design journal of the Discovery Institute.) Every indication we currently have from this kind of analysis, then, is that genetic data do in fact rule out a single pair of ancestors within the last 500,000 years. Such a date would put Adam and Eve well before the appearance of Homo sapiens, making them members of an earlier Homo species.

I have spent so much time talking about frequency distributions not because they provide the only or the best approach to reconstructing human history, but because it is intuitive and easy to visualize—at least compared to other approaches. A number of more sophisticated methods, methods that extract more historical information from genetic data, have been developed and used to probe the history of our species. I will touch on one class of these approaches more briefly, just to give some idea of what’s out there.

These approaches do not pool information from the entire genome together, as the frequency-based ones do. Instead, they make use of the fact that every segment of the human genome has its own genealogical tree, one that differs from those of neighboring segments because of ongoing recombination, that “mixing” of bits of the genome I mentioned earlier. The tree for each segment can be reconstructed based on the genetic variants present in different people’s genomes; the reconstruction is both complex and computationally intensive. The length of each branch of the tree (measured in generations) can then be estimated from the number of mutations that have accumulated along it. In a second step, the size of the ancestral population can be inferred from the shape of the tree. For example, a small population is signalled by many branches coming together during a particular period, indicating that there were a small number of ancestors during that time.

Using this technique, researchers have reconstructed the size of ancestral human populations, working back in time for various modern groups from around the world. What they find is very similar to what we saw above: non-African populations all show signs of a substantial bottleneck tens of thousands of years ago, while African populations look like they have always been large, stretching back at least hundreds of thousands of years. For example, one study from 2019 estimated an effective population size for the ancestors of Finns that was as low as 2000 at one point, while the minimum sizes for African populations varied between 8000 and 12,000, depending on the population.

How certain are the results of these more sophisticated methods, particularly in the face of created diversity? One researcher, Joshua Swamidass, a professor at Washington University, has looked at this question in depth. He looked at the number of distinct reconstructed genetic lineages for each DNA segment. In the present, there is one lineage for every sampled genome, and that number shrinks as you look further into the past, as lineages join at their common ancestor. The key point: as long as there were more than four lineages, there must have been more than two individuals around to carry them. His conclusion is that a single original couple would have had to have lived at least half a million years ago. With this very different approach, then, we find that the data points to a similar answer as before.