Reply to Moran: The Grandmother Hypothesis of Menopause

In my earlier post I discussed the "Grandmother Hypothesis" as an explanation for human reproductive senescence, or menopause. A problem arises in understanding why women forgo one-third (and sometimes as much as one-half) of their reproductive lives, a condition unique in the natural world. Could this just be a neutral mutation, an artifact of longer human lives, or might it be a product of natural selection? If the latter, what selection pressure(s) could result in this unique human adaptation? The grandmother hypothesis posits that women who stopped ovulating in their golden years were freed from the costs of reproduction and were better able to invest in their existing children and grandchildren (thus helping to ensure that more individuals with their menopause inducing genes thrived and had children themselves).

University of Toronto Biochemist Larry Moran at Sandwalk takes issue with my support for this hypothesis, as discussed by Razib at Gene Expressions. Moran refers to the hypothesis as an "adaptationist just so story" and insists that the study cited offers no evidence that menopause has adaptive value:

Given that there were many families that received no help from grandmothers, whether they had the menopause allele or not, and given that there were many families who received help even if the grandmothers did not have the menopause allele, the question is "what is the adaptive value of menopause under those circumstances."

Moran asks an excellent question and it touches directly on the problem of testing hypotheses of human evolution. How do you test for the preferential selection of allele frequencies if a trait is already fixed in a population? Unlike in mice or fruit flies it's impossible, for both ethical and practical reasons, to engineer human knockouts that don't develop the menopausal phenotype in order to conduct a controlled experiment. However, the grandmother hypothesis was given strong support by a 2004 study in Nature by Mirkka Lahdenperä, Virpi Lummaa, Samuli Helle, Marc Tremblay and Andrew F. Russell. As they state the problem:

Previous research on humans provides some evidence that post-reproductive mothers can benefit the reproductive output of their offspring. However, it has not yet been possible to test whether prolonged post-reproductive longevity in humans is associated with greater grandchild production, and hence greater fitness, because data covering the complete reproductive histories of several generations of individuals are scarce.

Using multigenerational data sets from the eighteenth and nineteenth centuries (prior to modern medicine) they conclude that the fitness benefits are substantial:

Grandchild survival to adulthood is enhanced by 12% when grandmothers are under 60 at their birth, but by only 3% when grandmothers are over this age. Grandmothers have no effect on the survival probability of their grandchildren between birth and two years of age (GLMM: 21 = 0.08, P = 0.78), but have significant positive effects when offspring are between two and five (GLMM: 21 = 7.09, P = 0.008), and between five and 15 (GLMM: 21 = 4.18, P = 0.041).

(For those who don't understand what these numbers signify, the important number is P. Any P-value at or below 0.05 means the probability meets the strict threshold required of modern science.)

So, in other words, post-menopausal women had a huge impact on the survival of their grandkids. This left Moran impressed, but perplexed:

12% is, indeed, a very large effect but what does it have to do with evolution?

What Lahdenperä et al. have done is demonstrate that post-menopausal women actually do show significant fitness benefits. As I'm sure Moran knows, fitness benefits are the precondition for evolution. But the story doesn't end there.

Following up on Lahdenperä's work, Daryl P. Shanley and colleagues writing in the Proceedings of the Royal Academy: Biological Sciences utilized a remarkably complete and instructive data set from the Keneba and Manduar regions of Gambia. They were able to locate birth and death records that date from 1950 to 1975 and include 5,500 people. Furthermore, since the records predate the existence of a medical clinic they offer a window into the world of hunter-gatherers without the benefits of modern health care. What the data reveal once again is that children were significantly more likely to survive to adulthood if they had a grandmother's assistance.

Their results show children were more than 10 times less likely to survive if their mother died before they reached two years old. However, children between one and two were twice as likely to survive if their maternal grandmother was still alive. This was the only other relative that mattered (sorry dads - I'm a new father myself so I feel your pain).

Shanley et al. then took their conclusions to the next level by modeling the fitness benefits that would occur if women experienced menopause later or earlier than age 50 (as women do now worldwide). According to their data the Gambian population had a growth rate of r=0.013 (derived from Table 1 above). The authors then calculated how r would vary if they imposed menopause at different ages between 45 and 65.

Estimated relationship between the intrinsic rate of natural increase, r, and age at menopause for four cases: (i) no effect of mother or grandmother on child survival; (ii) death of mother assumed to result in 13.4-fold increase in mortality of children aged 0-1 years and 11.7-fold increase in mortality of children aged 1-2 years and death of grandmother assumed to result in 2.0-fold increase in mortality of grandchildren aged 1-2 years; (iii) as (ii) where the 11.7-fold increase in mortality of children aged 1-2 years following the death of the mother is assumed to decrease linearly between ages 2 and 15 years; (iv) as in (ii) where the the 2.0-fold increase in mortality of grandchildren aged 1-2 years following the death of the grandmother is assumed to decrease linearly between ages 2 and 15 years.

Their conclusions are that the women experience the greatest fitness benefits by undergoing menopause right around age 50. Too early and they forego further reproductive opportunities. Too late and they actually have reduced fitness by continuing to have children and not assisting their daughters and grandchildren. Menopause, it seems, was a fitness enhancing solution. These two studies have gotten around the problem of setting up a controlled test using human subjects and developed an empirical means to test the fitness benefits of reproductive senescence.

Moran's concern that the grandmother hypothesis is a "just so story" is an important one. Too often, researchers construct an argument that sounds good but that can't be tested empirically. While there's certainly room to critique these studies based on the reliability of the data used or the model Shanley et al. develop, the fact is that the grandmother hypothesis is on solid empirical ground. I would encourage Moran to take a close look at the numbers himself to see if they meet his standards.

However, Moran has a further question about the evolution of menopause:

Since you are a supporter of this adaptationist explanation can you describe for me the kind of society where you think this allele became fixed in the population? Was it a hunter-gather society of small bands or a large agricultural society of small towns? Or something else?

Unfortunately there's no evidence to offer Moran an informed answer, but there may be a way to find out. A 2004 study in The Lancet found that premature menopause is associated with mutations in mitochondrial DNA. Since mtDNA is passed from mother to daughter it is not recombined the way nuclear DNA is during sexual reproduction. This makes it an excellent genetic clock (I'm sure everyone has heard of "Mitochondrial Eve"). It's possible that a similar global analysis of mtDNA could be performed in order to date this potential marker of menopause. Once an approximate date is established paleoanthropological data can be used to estimate group size and foraging patterns at the time menopause may have been undergoing selection. It's a place to start.

In conclusion, the grandmother hypothesis is far more than a mere "just so story." Hypotheses that avail themselves to testing and the potential for disproof are the markers of good science. It's often as easy to throw out the accusation of a just so story as it is to make one. Fortunately, by engaging with the evidence used to support a given hypothesis, scientific progress can be made.


ResearchBlogging.orgLahdenperä, M., Lummaa, V., Helle, S., Tremblay, M., & Russell, A. (2004). Fitness benefits of prolonged post-reproductive lifespan in women Nature, 428 (6979), 178-181 DOI: 10.1038/nature02367

Shanley, D., Sear, R., Mace, R., & Kirkwood, T. (2007). Testing evolutionary theories of menopause Proceedings of the Royal Society B: Biological Sciences, 274 (1628), 2943-2949 DOI: 10.1098/rspb.2007.1028


More like this

From Lahdenperä et al., how was the data analysed? The p values are not very low, and the summary implies they may have utilized multiple testing of the same cohort. Any multiple-testing correction could probably make an 0.008 fail, depending on the number of tests.

I can't access the nature article from home...

Is it reasonable to expect that enough early humans survived past 50 to make menopause all that beneficial?

By ponderingfool (not verified) on 13 Oct 2009 #permalink

Larry sees an adaptationist just so story behind every .... trait.

I can't access the study or the report (without paying money) so I can only propose some different hypotheses.

I have no particular axe to grind about 'just so stories' but I do think we tend to be too easily satisfied by simple stories which fit our preconceptions in a tidy way.

Over on Larry's blog I wondered if the menopause was a byproduct of the evolved trait to limit the number of children a woman could have to an optimum size.

Reading a little more about the data over on this blog, I still have a question about the grandmother hypothesis. Has the original work tried to identify how the grandmother makes a difference to child survival? For instance having a living grandmother may correlate with more children surviving, but may not be wholly the cause of their survival. Any woman has on average half of her own mothers genes. If the grandmother survived beyond 50 because of 'long lifespan' or 'defeat illness' genes, these would be passed on to her grandchildren too, explaining their better survival rates. The favourable genes could be sex linked or there could be some epigentic factor to explain why male careers show no correlation with child survival. I wonder if the original data can be read another way?

By DiscoveredJoys (not verified) on 14 Oct 2009 #permalink

I would like to hear the response to DiscoveredJoys' proposed alternative explanation.

@Jim: Here's the statistical methodology from Lahdenperä et al:

Statistical analyses were conducted using SAS (SAS Institute Inc., release 8.02, 1999â2001). Relationships between a female's post-reproductive lifespan and the number of grandchildren born to her were analysed using general linear models, in which number of grandchildren born was fitted as the response term to a normal error structure and female longevity was fitted as the main fixed effect. In these analyses, we restricted our data to include only those women who lived past menopause (n = 339 for Finland and n = 2,362 for Canada)âthe age at which 99% of females had finished reproducing (49 yr). The analyses control for socio-economic status and geographic differences in living conditions (population)19 in Finland and temporal differences in living conditions (birth cohort in both countries).

Effects of post-reproductive mother presence on fitness correlates (Fig. 2aâc) and underlying life-history traits of offspring (Fig. 2d, e) were analysed using GLMMs, which allow both fixed and random terms to be fitted to a model, with random terms controlling for repeated measures within mother28 (and grandmother in Fig. 2e). Satterthwaite's formula29 was used to approximate the denominator degrees of freedom of each fixed effect. In these analyses, the response terms were fitted to a normal error structure and mother presence/absence was fitted as the main fixed effect. We defined presence/absence as whether a post-reproductive mother was alive versus dead (Fig. 2a, b, d), or living in the same village versus elsewhere (Fig. 2c), when each offspring started reproducing. In Fig. 2e, the presence of a post-reproductive (grand)mother was defined as whether or not she was alive when the grandchild was born. We controlled for confounding effects (where appropriate) of: mother's age, offspring's socio-economic status, population19, birth cohort, sex, birth order, the number of competing siblings, the survival status of the directly previous child born in the family and the repeated measures of the same mother.

The effect of post-reproductive grandmother presence on grandchild survival was analysed using log-rank test and survivorship functions estimated using the KaplanâMeier method. To investigate the period when the presence of the post-reproductive woman was most important for the survival of grandchildren, GLMMs were performed, where the response term for whether or not the grandchild survived to a certain age was fitted to a binomial error structure and logit-link function. Analyses controlled for socio-economic status of a parent, population, birth cohort and order, sex of grandchildren and repeated measures within parents.

In all models, significant model terms were determined using the backward elimination technique, and subsequently checked using the forward technique. The residuals of all models were normally distributed (GLMs, GLMMs) and the variances were equal (GLM: Levene's test P > 0.05). All statistics are from the GLM unless otherwise indicated.

ponderingfool: Is it reasonable to expect that enough early humans survived past 50 to make menopause all that beneficial?

Why would you wonder that? For all the data I've seen, pre-agricultural societies have long lifespans -- once you eliminate deaths under five. The numbers we get usually include infant mortality -- but since the distribution isn't normal, the "average" lifespan is a completely useless metric.

As I understand it, you have a huge hump of deaths between 0-5 (which is why some societies didn't even name children till they were at least one). Then there's another hump of deaths in 20's -- warfare and first-birth deaths. But if you make it past there -- you had the same lifespan we have today.

Our idea of short lifespans for our forebearers comes from heavy duty agriculturialists (where most humans were used basically as energy sources, completely expendable) and industrial societies (where most humans were used as screwdrivers -- again, completely expendable). Our current lifespans are actually a return to most of human history, minus the infant mortality.

@DiscoveredJoys: This isn't just about survival rates, this is about fitness rates:

In the presence of a living post-reproductive mother, both sons and daughters have a greater number of offspring in their lifetime (Fig. 2a), and raise more to adulthood (age 15) (that is, greater lifetime reproductive success) (Fig. 2b). Additionally, a living post-reproductive mother has a significant positive effect on her offspring's individual lambda (GLMM: F1,524 = 4.22, P = 0.04). Finally, the longer a post-reproductive mother survives after the offspring started reproducing (controlling for mother's age) the greater her offspring's lifetime reproductive success (GLMM: F1,281 = 8.42, P = 0.004).

I can't reproduce images here, so for Figure 2 please go to The Primate Appendices.

As for your question about if the data could be read another way, you'll probably want to contact the author's about that to see about accessing their original data. They're clear in their paper that it is grandmothers helping that results in increased fitness. But perhaps you can find something that they didn't.

@frog: You're absolutely right. Unfortunately, Moran may have made the same false assumption about average lifespan in his first post on menopause:

The most common just-so story is called the "Grandmother Hypothesis." It imagines a time in the distant past when humans females were fertile until the day they died, which for 75% of women who survived childhood was before the age of 30 according to a recent study.

I'd be curious to know what study he's referring to.

To further stir the pot, any good data on social animals with multiple generations remaining fertile? I'm thinking more on the potential disadvantage to a daughter of having her own mother competing with her as they both raise infants. Also, infanticide of daughters offspring by dominant females is seen in primates.

If more children with disabilities are born to older women won't it make sense for reproduction to stop?

Also, wouldn't older women be more likely to die at childbirth or sooner after than younger women?

Without mothers the children of older women would be less likely to survive.

No, older women are not more likely to die at childbirth, certainly not those who have already successfully given birth. The first birth is the most dangerous.

Yes, older eggs are less viable. But let's be very clear. The question is not that you can't invent environments where menopause yields an advantage, you can. The question is, are they historically significant?

And if you observe primates, what do you actually observe over what you imagine grandmothers do? I say primates because menopause is NOT uniquely human.

ej: "If more children with disabilities are born to older women won't it make sense for reproduction to stop?

Not from the perspective of that individual. However many children they've had up to that point, if they can have more (even if, statistically, those additional offspring are less likely to mature and reproduce themselves), then it is an additional fitness advantage for that person to be able to continue to reproduce.

That is the heart of the issue: why have menopause if, at first analysis, it lowers fitness. One possibility is that there is a net fitness advantage, such as improving survival of grandchildren. Or maybe the fitness advantage is so little that drift was able to bring in menopause.

In any event, the Larry Moran view** is not really helpful.

**In a nut shell, the Larry Moran view is: "I don't know, and probably never will, so there is no reason for anybody to investigate it, and if you propose a hypothesis (even *if* that hypothesis makes predictions you then test; (i.e., you do *science*!)), then you are an idiot adaptationist making up "just-so" stories, and so you and your research are fair game for any and all ridicule".

A rather creative interpretation of scientific criticism there Divalent. In particular, I missed the part of Moran's comments that imply one should not investigate such thing as menopause.
Not making unsupportable conclusions seems distinct from not investigating. And directly to the specific issue, how does one establish an adaptive advantage without being specific about the environment in which selection supposedly occurred?
It seems to me that the scientific nihilism is worry about is that which attacks those who ask question or shows fear of having to conclude "we still don't know, and may never know".

I have a vague recollection of a paper by Sokol(?) back in the 70's(?), which showed that the most important factor in population increase was not number of offspring per female, but rather age of mother at first birth. As I recall, a population with first birth at 15, then second at 16, and third at 17, then no more, would grow faster than a population where first birth was at (20 or 25?) and each mother had as many children as possible.

If so, then the fitness of having children early (with a grandmother to help out?) is perhaps greater than the ability to have children at an old age.

By Jim Thomerson (not verified) on 16 Oct 2009 #permalink

The grandmother hypothesis posits that women who stopped ovulating in their golden years were freed from the costs of reproduction and were better able to invest in their existing children and grandchildren(thus helping to ensure that more individuals with their menopause inducing genes thrived and had children themselves).

So if I'm understanding this, the adaptative mechanism is such that menopause-inducing alleles are transmitted from mother to daughter during procreation, but this transmission seems to be occuring only because of a sociocultural adaptation that works such that grandma's support allows female grandchildren to survive and reproduce.

It's possible that rather than menopause-inducing genes, what was/is transmitted is not menopause-inducing genes, but longevity and/or resiliency genes. There's a study here showing that a relatively late menopausal onset predicts a longer lifespan. This seems to be the opposite of what one would expect if the adaptive function of menopause is to stop women from having their own kids so they can focus on the grandkids. That is, there should be a prediction of a longer lifespan with an early menopausal onset, if it is true that "women who stopped ovulating in their golden years were freed from the costs of reproduction".

So based on that data point, my guess is that longevity/resiliency genes are what was/is being passed down and menopause is an artifact of increased longevity. This might explain why on virtually every country on the planet, women live from 4-7 years longer.

A couple of questions such a hypothesis raises: Has the gender difference in mortality always existed, or has this difference emerged over time? Is there a difference in the average lifespan of grandmothers who supported vs. those who did not support, which accompanies the 12% finding?

Grandchild survival to adulthood is enhanced by 12% when grandmothers are under 60 at their birth, but by only 3% when grandmothers are over this age.

This could be because grandmother's may be less capable of helping because of greater physical, mental, or social incapacitation with increased age.

Grandmothers have no effect on the survival probability of their grandchildren between birth and two years of age (GLMM: 21 = 0.08, P = 0.78), but have significant positive effects when offspring are between two and five (GLMM: 21 = 7.09, P = 0.008), and between five and 15 (GLMM: 21 = 4.18, P = 0.041

This seems to support the longevity/resiliency gene transmission hypothesis and weakens the hypothesis of the transmission of menopause genes, and suggests what's happening may be a nature-nurture interaction.

If grandma support is the key factor driving survival of grandkids into adulthood, it does not make sense why the effect should start at 2. Why does the effect not start right away? That suggests some sociocultural phenomenon is at work.

On the other hand, the finding might suggest the possibility of inheritance of resiliency/longevity genes in the maternal line. That is, the grandma effect does not occur from 0-2 because (perhaps) of genetic factors causing birthing/fetal problems, possibly leading to high mortality rates. However, those possessing the longevity/resiliency genes may be the ones that survive past age 2, and who get taken care of by grandma.

The longevity/resiliency genes are then passed on in the maternal line (influenced by grandma support), and the residual effect would be menopause accompanied by a longer lifespan.

If true, it should predict an increase in the lifespan of women (relative to men) over time, that we now observe in nearly every country on the planet.

By Tony Jeremiah (not verified) on 16 Oct 2009 #permalink

Very thoughtful response. Larry loves to criticize and ridicule without really thinking things through.

Additionally, this study suggests that the physiological mechanisms that prolong the lifespan and delay menopause according to this study , are the same ones that increase the risk of birthing complications in women who decide to have children relatively late.

Assuming those data are replicable and are indicative of evolutionary mechanisms associated with reproduction, there's some suggestion that the function of these mechanisms are to ensure childbirth happens earlier (probably optimally by age 35), prolong lifespan and delay the onset of menopause.

Again, these studies seem to be pointing toward the inheritance of longevity/resiliency genes in the maternal line, with menopause as an artifact of enhanced longevity.

By Tony Jeremiah (not verified) on 17 Oct 2009 #permalink