Evolving ideas

A recent post on the Gene Expression blog (here) alerts me to an article on selection for human longevity in PLoS One (here).

Explaining ageing is actually harder than it might seem and there is no particular reason to suppose that ageing should occur at all. One explanation, described by Peter Medawar, is that genes with deleterious effects later in life are relatively less effectively selected against because of a general increase in mortality with age (the longer you live the higher the chance of accidental death) and can persist in a mutation-selection balance. Decreases in the intensity of selection with age can also favour genes with beneficial effects earlier in the lifespan and deleterious effects later. This scenario, first described by George Williams, is called antagonistic pleiotropy and can take a number of forms depending on the sorts of trade-offs lying behind these gene effects.

But there is a problem which is unique to humans (ed. it turns out I am wrong that this is unique see here!) and that is the female menopause. As Hamilton pointed out there should be a sudden increase in mortality after the age of menopause in both sexes because an expected fertility of zero effectively negates selection against deleterious genes acting at this time. Arguments aimed to try to fell this “wall of death” are generally focussed on intergenerational transfers and the best known of these is the “grandmother” hypothesis which suggests that the continued presence of post-reproductive females can enhance survivorship of grandchildren and other relatives through the provision of resources and time. In this way longevity is kin selected.

But the PLoS One article by Tuljapurkar et al. offers a new explanation, which is that male reproduction past the age of female menopause can strengthen selection for longevity and assuage the rise to infinite mortality predicted by Hamilton. Collating data from several cultures they show that males continue to show non-zero fertility rates after the age of female menopause (which is not surprising given that males tend to mate with younger females across the board), but also they show that the shape of the fertility function is different in males compared with females. It is not simply an echo of female fertility rates at younger ages, but instead shows an extended tail of male fertility with increasing age. This effect is likely a result of several factors including serial monogamy (wherein males are more likely to re-marry than females), polygyny (in which males have multiple partners) and the existence of high status males (who show persistent sex appeal into old age). It is important to remember that genes for longevity can be shared between the sexes so that theories of ageing should take account of fertility in both sexes. Tuljapurkar et al. explicity modelled the dynamics of fertility in both sexes to produce a model supporting their conclusion that male age-specific fertility rates can explain the persistence of human longevity past the menopause and the absence of a “wall of death”.

This week’s Science magazine has an interesting study (here for summary) about the relationship between parental investment and cooperative breeding in the superb fairy-wren. Sometimes these birds breed in pairs and sometimes cooperatively.

In cooperative breeding (sometimes unrelated) individuals assist a breeding pair by provisioning their young. It turns out that spotting the benefits of this has been quite hard: the chicks are given more food but they don’t seem to end up any heavier for it. Russell et al. nicely show, with general linear models, how the beneficial effects of increased provisioning are hidden by a corresponding decrease in maternal investment in the egg. It turns out that in cooperatively breeding situations mothers produce lighter eggs with less yolk and this is compensated for by the increased provisioning. This implies that the benefits of cooperative breeding are not to the chicks but to the mothers who may be able to conserve resources for future reproductive attempts. Even more elegantly, Russell et al. reciprocally cross-foster the chicks from pairs to groups and vice versa showing that offspring from eggs hatched in groups and raised by pairs fare particularly poorly.
An interesting and more openly speculative part of the paper is the authors’ explanation as to why it might be that cooperative breeding occurs at all. Apparently delayed reciprocity (whereby helpers specifically receive benefits later) is unstable but benefits accruing directly from augmentation of group size (which may include helping, but also things like predator vigilance) may be the culprit. The authors cite this paper in support. Now male birds tend to disperse less from the natal area than females so we expect males to benefit more from group augmentation and Russell et al. measured increased provisioning to chicks with increasing male numbers. Group augmentation seems to me a fascinating explanation for altruism in groups and this study strikes me as a superb (sorry!) example of the effective testing of adaptive hypotheses.

I’m thinking about the relationship between induction and parsimony. Parsimony is the principle that urges us to adopt the theory which makes the fewest assumptions, when faced with a choice between two or more theories that all equally well explain the evidence presented. One way to understand why this might be justified is that theories which make fewer existence claims are responding better to the evidence which is not presented. This assumes that the absence of evidence for an existence claim is, to some extent, evidence for its absence. The non-presence of some evidence therefore favours the simpler theory (making fewer existence claims) over the others. The problem with this is the problem of induction. We cannot be certain that we are not making an inductive fallacy (roughly speaking assuming that the future will be like the past). Now I said above that absence of evidence is “to some extent” evidence for absence. This indicates that a probabilistic view of parsimony may also be adopted. If a theory postulates the existence of two things each with some probability (independent of the other) then it has a lower probability (multiply them!) than a rival theory which postulates the existence of only one of these. This seems to me to be true regardless of our uncertainty about what sorts of probability distribution underlie reality (although these will affect the usefulness of parsimony). Am I slipping up here or repeating myself? Finally the interesting exception to this is when the prior probability of particular existence claims is increased by the theory within which we labour. I believe evolutionary theory increases the likelihood that developmental mechanisms will be complex. This is simply because we must consider them as products of descent with modification (whether this is achieved by genetic drift or natural selection).

Natural selection is the most important process in evolution.

The meaning of this statement depends on how importance is judged and on what is meant by evolution. So, for precision, what I will actually defend in this posting is the proposition that natural selection is always necessary, if not sufficient, to explain adaptive evolution, by which I mean change between generations which improves the fit between an organism and its environment.

This statement is general because it applies to as yet unexamined cases. So it might seem to be an anti-empirical claim. It is not meant that way; data should be collected and hypotheses tested (as it already has been in support of evolution by natural selection in particular instances). It is just that whatever is revealed about the evolutionary processes leading to adaptation in any particular case will, I believe, always evidence or logically invoke natural selection in a non-trivial role. I would like to contrast this “strong” position with a “weak” one that says that natural selection cannot be credited with an essential role until this evidence is collected. It needs stressing at this point that evolution (in the sense of descent with modification) is massively evidenced across the board (biogeography, fossil record, vestigiality, sequence data, genetic codes, experimental evolution, etc.) and that’s not what’s at issue here.

What kind of objections can there be to this? Obviously creationists (also read “intelligent design theorists”) will not agree with me – mostly, of course, because they are committed to and trying to justify a particular theistic worldview that is plausible only if one is ignorant of the facts supporting evolution in general. Such pretend rationalism is so much fluff and anyone interested needs to learn a lot of relevant facts and should buy a decent textbook on evolution rather than reading blogs (try this). But, as we shall see, my specific proposition might seem to be at odds with some scholarly opinion.

It is likely that natural selection plays a trivial role in the evolution of eukaryotic genome structure. This is the core argument of a recent paper by Michael Lynch published in the Proceedings of the National Academy of Sciences USA. (Note that Lynch has written this paper also and a book reviewed here). Long-understood, but often-overlooked non-selective processes better account for the emergence of higher level properties such as (genetic) complexity, modularity (in development) and evolvability (the ability to respond to selection and generate novelty). Lynch’s polemical point is that there are only four major evolutionary “forces”: mutation, recombination, genetic drift and natural selection, and that the field of population genetics already provides all the mathematical tools needed to analyse the relative significance of these. (Nothing is said about migration’s impact). Paying particular attention to genetic drift he emphasises that these tools should be regularly used to test non-adaptive explanations.

Genetic drift is akin to sampling error in statistics. It tends to occur in small populations of organisms when each generation is constituted from a small sample of the gametes produced. The average proportions (or “frequencies”) of gene copies (or “alleles”) in small samples tend to deviate by chance from those in the population from which they are sampled. In this way a new mutant allele can be lost from a population or can increase in frequency so that all individuals carry it. Genetic drift can be strong enough to scupper natural selection if the fitness benefit of a new allele is small and therefore not instantiated in differential reproductive output. The key point for Lynch is that the population size (or more technically the effective population size) is small in multicellular eukaryotes and this permits the accumulation of genetic novelties such as introns (non-coding segments located within gene-coding regions) with mildly deleterious phenotypic consequences (considered individually). From this perspective Lynch takes a swipe at Dawkins whose “agenda to spread the word on the awesome power of natural selection”, at the expense of other evolutionary mechanisms, represents “a view that is in some ways profoundly misleading”.

Now being a theoretical pluralist, having an open mind about the relative role of all evolutionary processes as Lynch proposes, sounds like a good idea. But my point is that, where adaptive evolution is concerned, we can expect a role for natural selection. This suggests that, taking into account his didactic role, we may at least partially forgive Dawkins. The existence of adaptation is striking and invites public curiosity. Unfortunately it also triggers the intuition that a system is designed. So there is an emotional reward in understanding how adaptation works and sometimes work to be done to overcome biases in intuition with the relevant facts. It is no use telling people that most evolutionary change is not adaptive. That is even though it may be true.

Now to defend my main point that natural selection is important for adaptive evolution. Lynch correctly points out that “the effects of mutation and recombination are nonrandom” and that “by magnifying the role of chance, genetic drift indirectly imposes directionality on evolution by encouraging the fixation of mildly deleterious mutations and discouraging the promotion of beneficial mutations.” This is true and makes talk about selection for increased complexity loose talk, but these non-adaptive forces are either random with respect to organismal fitness or detrimental to it. So any explanation of improved fitness and of adaptation to the environment will rely on a process that can discriminate in favour of fitter variants (see my posting on Lamarck below) and this is natural selection. The truth of the weak and strong versions of my position, i.e. whether or not we can always expect natural selection to underlie adaptation, depends on our setting a threshold for the improbability associated with adaptive change in general. If adaptation is unlikely to occur by chance then we can conclude a priori that natural selection is likely. Let me stress again that it is always worth doing the experiments to show it. In particular experimental evolutionary work is useful because it breaks down the process of adaptation by probabilistic steps (and shows why it is not a mere tautology).

Now the main problem with my position so far is that it is not always easy to know whether or not something is an adaptation, let alone assess it’s improbability under non-selective scenarios as Lynch does. Probably the best-known criticism of adaptationism was made in Gould and Lewontin’s 1979 famous spandrels paper. Some of the literary allusions in this piece have taken on a life of their own since and it common to hear criticisms of adaptive hypotheses to the effect that they are “just-so stories”. Perhaps less common is the assertion that the adaptationist programme is a result of a “Panglossian” imagination. This is one that sees all phenomena as explained by their current function, i.e. one in which adaptation is presumed to be omnipresent. (This is like Dr. Pangloss, a character in Voltaire’s Candide, who thought that everything in this world was for the best and who constructed tenuous explanations to fit his supposition – a lampooning of theology).

When these criticisms are merely asserted they are, in my view, patently unfair. Many adaptive stories are subjected to testing (Lynch is adding value here) and the attractiveness of adaptive explanations cannot be taken to automatically disqualify them. To be fair to Gould and Lewontin, they did not just make these assertions. Instead they suggested “a partial typology of alternatives to the adaptationist programme”. Genetic drift as a source of non-adaptive evolutionary change was discussed although the paper’s thrust was morphological evolution. Lynch seems to be expanding on this for the molecular age by marrying molecular biology and population genetics. The take-home point is that these tools can be more widely applied to tighten up the science.

So far so good, but where does Lynch’s thesis take people who are thinking seriously about evolution and behaviour? Will it always be impossible to find support for natural selection on complex behavioural traits without first identifying the responsible genes and then analysing them using appropriate algorithms or at least gathering relevant population data? Can we make any general statements about the relative role of natural selection on traits “higher up” the phenotypic scale?

Well I reckon there are good reasons to think that most change at the genetic level is due to drift (chief among these is small population size in multicellular organisms as noted by Lynch) and as described there are good reasons to think that most adaptive behavioural and morphological change is due to selection. This seems like an interesting paradox: that as we move “up” the phenotypic scale the prior probability of natural selection increases, although of course it is perfectly reasonable to suppose that most visible changes are underlain by only a subset of changes at the genetic level. An alternative interpretation is that traditional biologists and evolutionary psychologists tend to find interesting things adaptive (the Panglossian tendency). I submit, instead, that they find adaptive things interesting and tend to ignore non-adaptive things like having precisely five fingers. The difference between scientists working at different levels of the phenotypic scale is then simply one of what interests them rather than something fundamental about evolution at different levels or which side of some great theoretical schism they fall.

I also think that arguments about complexity, modularity and evolvability are distractions. I agree with Lynch on these in that I think the prior probability of selection is low for these too (and perhaps we can fault Dawkins here). (Any thoughts on modularity of mind here – or is this too tenuous a link?)

So to sum up – the only a priori plausible mechanism for adaptive change is natural selection (possibly there are self-organising properties at lower probabilities) and discussions about how significant it is should take this into account the centrality of adaptation in the field under review. As good scientists though we should seek independent evidence for natural selection and try to exclude more chancy processes. And we should be wary of labeling something adaptive when it is not. Looking for evolution by natural selection we are thus between Scylla and Charybdis. They can each swallow a lot of sea water – but nobody need doubt the existence of a channel through or feel any compunction in charting this voyage for public edification and enjoyment.