Multi-level selection and The Selfish Gene

Yesterday, I attended a seminar run by the University's Institute for Complex Systems Simulation by Samir Okasha, a Professor of Philosophy of Science from the University of Bristol and author of the book "Evolution and the Levels of Selection" (among others). The talk was entitled 'Individuals versus Groups in Evolutionary Biology' and Prof. Okasha gave a very interesting presentation about some of the history and issues surrounding the discussions (and sometimes arguments) about "Group Selection" and its modern incarnation, "multi-level" selection.

It looks like an interesting book too and is on my ever-growing reading list. I'd particularly like to ponder some more his thoughts on emergent group properties - something I do not currently have the time, space or philosophical nonce to explore further in this post.

There was one key aspect of the debate that, in the interests of time, was not covered in detail in his talk: the issue of heritability and what that means for "Units of Selection". The more I think about it, the more I think it is a real barrier for people understanding the problem and, in my opinion, leads to all sorts of confusion about how evolution and selection work.

This is a quote from an Amazon review of his book that sums up the key issue quite nicely:

So often we are bombarded with 'scientists' giving us their metaphysical views as if they were 'scientific fact'. It is therefore refreshing to find a philosopher looking at a science and seeking to clarify the various concepts in that science.

Okasha observes that the various life forms are arranged in a hierarchy:
Ecosystems
Species
Colonies
Organisms
Cells
Chromosomes
Genes.

Generally reproduction occurs at the same level in the hierarchy: organisms reproduce to give organisms; chromosomes divide to give chromosomes; colonies divide to give colonies, and so on. According to the logical formulation of the theory of `natural selection' a) variation, b) differential fitness (different rates of survival and reproduction) and c)heritability (parent - offspring correlation) are required to produce evolutionary change. All these may be present at each of the levels in the hierarchy so there is nothing that necessarily restricts selection to any one level, say at the level of the gene. To claim that selection always occurs at the level of the gene is to confuse the result of selection (the proportion of the various genes in the gene pool) with the process of selection (where in the hierarchy the winnowing actually occurs). [My emphasis]

This is an argument that I have come across a few times on internet forums and like - often by non-biologists. (I'm not sure why the reviewer puts quotes around 'scientists' - perhaps this is an unfair dig at Dawkins. When these arguments appear, they are often accompanied by a barrage of anti-Dawkins nonsense about dogmas and how our old, flawed understanding of evolution is being overthrown etc. At best, this is a gross exaggeration. In my opinion, it is utter hogwash.)

Quite simply, I don't think this argument works because it overlooks something very important. I have highlighted the key phrases in bold. This review has the matter utterly backwards. To say that selection is occurring at a level other than the gene and not the gene (and "gene" in this context must have the correct evolutionary meaning not the biochemical meaning) is to confuse the agent of selection, which can be gene, cell, organism, family, whatever, and the target of selection - the "gene". This is because, for selection to work, there has to be heritability and this heritability is not simply "parent - offspring correlation".

(At this point, I would like to make it clear that I do not think Samir Okasha makes this mistake. I've not read his book yet but in his talk he was very clear to make the distinction between causality in selection - what we call direct and indirection selection, which correspond to causal and correlative changes in gene frequency. He also pointed out that there is no conflict with multi-level selection and "The Selfish Gene".)

For selection to work, there has to be a causal link between the heritable trait and differential fitness. Mere correlation is not enough. It is enough for evolution - there will be a change over time - but it is not enough for natural selection. And this is where genes are special. Yes, selection can potentially act at some of these different levels - the collective properties of the family, tribe, species or ecosystem can affect the fitness of the genes therein - but only the genes make copies of themselves. Only the genetic information is passed on - all of the physical aspects - the DNA, the chromosomes, the cells, the bodies, the tribes, the ecosystems - are transient vehicles for this information. Only if this genetic information gives rise (in an appropriate background) to the trait that influences fitness - whatever the level that fitness is manifest - will that trait be heritable and selection happen. Reproduction in the important sense - heredity - does not occur "at the same level in the hierarchy".

Has anyone actually demonstrated non-genetic inheritance of any higher-level trait? I'm not aware of any and whenever I have raised this in online discussions, I am normally just met with a barrage of anti-Dawkins nonsense or some vague notions about epigenetics, behaviour and "emergent" properties (which I advocate in general, by the way,) without any specific demonstration or model as to how these higher levels reproduce and pass on their traits to the next generation. Crucially, you have to do more that demonstrate that it could work mathematically or in a computer simulation - you have to demonstrate that there is a corresponding biological reality.

Which brings me to another important point. I would also question the notion of "fitness" at some of these higher levels. Ecosystems do not reproduce at all. There can be competition between groups of organisms, certainly, and long-term differential survival, which will result in evolution - just as random events such as floods and meteor strikes can influence long-term evolution through differential survival. But this is not selection. The ecosystem is changing because of individual success or otherwise and individual success is being influenced by the environment - the changing ecosystem - but an ecosystem is not directly spawning a new ecosystem that inherits its properties and goes off into the world to compete with different ecosystems. (It seems to me that there is one higher level entity capable of non-genetic inheritance - something championed by Dawkins himself. The cultural replicator, or "meme". This is not what multi-level selection is about, though, as far as I can tell.)

A final problem for non-genetic multi-level selection is that many of these "levels" don't really exist in a fashion that makes selection possible - they are part of continua rather than discrete entities. An ecosystem, for example, does not really mean anything specific. I am an ecosystem from the perspective of my gut microbes. The whole planet is an ecosystem. It is useful to drawn the boundaries at different points for specific study but we should remember that these distinctions are arbitrary. Even an "individual" is a woolly concept thanks to symbiosis - and we are probably all symbionts at the end of the day.

The only thing that is absolute is that you can break everything down to genes (genetic information) and their environment. The flow of information is one way. Genetic information is modulated - but not created - by the environment. (Even accounting for epigenetics, which modulates the environment but not the genotype, though this is for another post.) The Selfish Gene (and its Extended Phenotype) still wins.

Or does it...?

There is one problem that remains for the "Selfish Gene" and it is the same one that plagues almost all of biology. Just like all the levels above it, a "gene" (in the evolutionary sense) is just a mirage. In many ways, there is no such thing as a gene. There is just genetic information. We like to talk about a "gene for X" but really what we mean is "heritable genetic information that has a causal but environment-dependent tendency to produce X". This is just a problem of conception and language, though, not the underlying mechanism and theory. Selection is still ultimately acting on genetic information, and it is still selection at this level that gives rise to adaptations, but how you package this genetic material up into genes is, again, context-dependent and (thanks to recombination and mutation) can be complicated.

It fascinates me how we love to try and split continua - life, species, development, genes - into discrete packets even when no such packets exist and then tie ourselves up in knots because we can't let go of those arbitrary (and false) divisions we have made. Ultimately, I think the issue of Individual versus Group Selection might just be this problem, taken to another level.