Solving the Evolutionary Puzzle of Twinning

Twins have long been a source of fascination in human societies. Mythologies all over the world use twins to convey hidden messages, and literature repeatedly exploits their potential to amuse, or sometimes unnerve, an audience. But from an evolutionary point of view, they’re a puzzle. Twins have much higher mortality rates in early life than children born in single births and also pose a risk to the mother, increasing her likelihood of dying in childbirth. We seem to be a species designed to give birth to one infant at a time, like other large mammals. Why then has natural selection not weeded out the tendency for women to give birth to twins?

The answer may be that, while twinning itself carries risks, the underlying trait which leads to the most common type of twinning confers benefits. This trait is polyovulation – the tendency for women to release more than one egg from the ovaries each month. Dizygotic (or non-identical) twins result from the fertilisation of two eggs (‘zygote’ means fertilised egg), in contrast to identical (monozygotic) twinning, which happens when a single egg is fertilised but then dividing into two embryos.

Polyovulation confers benefits because it counteracts the surprisingly high likelihood of foetal loss. Not all the eggs a woman releases each month result in a successful pregnancy, even if she is actively trying to get pregnant. Some may not be fertilised, and most fertilised eggs don’t survive to full term. Exactly why isn’t clear, but this foetal loss may often be to do with chromosomal abnormalities which mean the foetus isn’t viable.

There is little data on exactly what proportion of fertilised eggs are lost, given that most seem to be lost before the woman herself is aware of the pregnancy. Identifying very early pregnancy loss requires monitoring of women’s reproductive hormones over time to pinpoint the hormonal changes which indicate pregnancy and then pregnancy loss. One of the rare pieces of evidence available, from a study in Bangladesh, suggests that very high proportions of foetal loss are common, particularly in older women: 45% foetal loss at age 18, rising to 92% by age 38.

Releasing more than one egg each month is therefore a way of ‘bet-hedging’: increasing the likelihood that at least one egg will result in a livebirth. Occasionally, more than one fertilised egg will make it through to term – and twins, or other types of multiple birth, such as triplets, result. So it’s polyovulation, the ability to release more than one egg at a time, which is favoured by natural selection, rather than twinning itself.

Another twinning puzzle: Why does the likelihood of twinning vary by maternal age?

This idea that twins are the ‘byproduct’ of an evolutionary advantage to polyovulation has been around for a while, but it doesn’t fully explain another twinning puzzle: its relationship with maternal age. Young mothers are relatively unlikely to give birth to twins; the chance of twinning increases with maternal age, but then declines again in the oldest mothers.

A recent paper I’ve contributed to, led by Wade Hazel and Joseph Tomkins, has just proposed a new explanation: that polyovulation is a ‘conditional strategy’, dependent on the mother’s age. This hypothesis assumes that young women produce only one egg at a time – their risk of foetal loss is low, so they don’t ‘need’ (in an evolutionary sense) to polyovulate; but at some point, a switch to polyovluation happens – older women have a much higher chance of foetal loss, so natural selection favours a switch to polyovulation to increase their likelihood of successful pregnancy.

Wade and his colleagues used mathematical simulations to test the plausibility of their idea. These models compared the reproductive success of different hypothetical strategies – a strategy of always single ovulating was compared with a strategy of always double ovulating, and with a conditional strategy of switching from single to double ovulation. The age at which switching happens was allowed to vary, to identify the best time to switch from single to double ovulation. This analysis showed that a conditional strategy, where switching to double ovulation occurred at around age 25, resulted in the highest reproductive success, in terms of leading to the most surviving children per woman by the end of her reproductive life (Figure 1).

Figure 1. Average number of offspring surviving to 15 years from simulations of the reproductive lives of 1000 women; the x-axis shows the age at which ‘conditional strategists’ (those who adopt a conditional strategy of first single ovulating then double ovulating) switch from single to double ovulation. The purple box indicates women who always double ovulate, yellow those who always single ovulate, blue those who adopt a conditional strategy.

In evolutionary terms, strategies with the highest reproductive success are likely to be favoured by natural selection, so that this analysis provides support for the hypothesis that polyovulation is a conditional, age-dependent strategy.  

This hypothesis is also able to explain why the frequency of twinning rises and then declines with maternal age. Younger women rarely have twins because they mostly single ovulate; double ovulation – and therefore the risk of twinning – increases as women age; but then at the oldest maternal ages, foetal loss is so high that, even with double ovulation, few embryos successfully make it through to term and the likelihood of twinning declines again.

What are the implications of twinning as an age-dependent, conditional strategy?

An age-dependent, conditional strategy of polyovulation could help explain why twinning rates are increasing in high income populations. Part of this increase may be due to reproductive technology, which often mimics polyovulation by implanting more than one embryo in women experiencing fertility treatment. But in high income populations women often delay first births until into their late 20s and 30s. If polyovulation is more likely in women in these age groups, then rates of twinning will be higher in populations where women have children relatively late, even without assisted reproduction.

These combined characteristics of polyovulation plus high rates of foetal loss also imply that many of us once had a twin. The ‘vanishing twin’ phenomenon, where a twin pregnancy detected early on subsequently becomes a singleton pregnancy, is well known in obstetrics; this new model suggests the frequency of this phenomenon is likely to be high, particularly among older mothers.

Twinning is a relatively rare event, varying from around 0.6 to 4% of all births, and this new paper suggests that its rarity is because it is an ‘accident’ arising from a strategy of polyovulation to counteract very high rates of foetal loss. Many questions still remain about this phenomenon: such as why foetal loss is so high, and why twinning rates vary around the world – because of differences in the likelihood of foetal loss, perhaps? But regardless of the explanation for the existence of twins, mythology, literature and the human experience is much richer because of this fascinating phenomenon.


Hazel, N. W., Black, R., Smock, R.C., Sear, R., Tomkins, J.L. (2020) An age-dependent ovulatory strategy explains the evolution of dizygotic twinning in humans. Nature Ecology & Evolution. DOI: 10.1038/s41559-020-1173-y