How fast does evolution happen?

... and other matters.

The following list represents widely held beliefs, either first order beliefs (things you hear people say) or second order beliefs (things implied directly by what people say):

1. Evolution is very slow.
2. It takes millions of years for a species to evolve into another species.
3. Evolution has stopped for humans.
4. Evolution occurs over "geological time scales."
5. Geological time scales are very long ... millions of years.
6. Geological and evolutionary time scales are similar to each other.
7. Evolution = Natural Selection.
8. "Macro evolution" is slow, "Micro evolution" is fast but tiny. Or whatever.

As a quick aside because I know you are thinking about this: There is also the theory of Punctuated Equilibria. This is where evolution does not happen in a lineage for a very long time, and then all of the sudden, lots of evolution happens. But still, even when speaking of Punctuated Equilibria, people assume and often explicitly state that the "sudden" evolution is fast for geological time scales but still something that you would not see ... it still takes a very long time. And, more importantly, evolution is seen as happening at different rates at different times with this model.

When you think about the above, you will find several conflicts and other problems. We'll get to some of those. But first, I want to propose a model of evolutionary change that I think is fairly accurate and at the same time useful for explaining and thinking about these issues. This is not a formal model that one would use in research (though if I formalized it, it would mathematically resemble several such models), but rather, this a pedagogical tool.

Imagine a room full of tables, and at each table there are several gamblers and they are all playing some form of poker. Each gambler has a fixed amount of money and when s/he runs out (by betting the last dollar and losing) s/he is gone and replaced by a clone of a nearby gambler.

If you want, you can imagine that each gambler has a different number printed on his or her green plastic visor, and then you can think of each gambler as an allele for a certain gene and keep track of them, but that is not important.

All we really need to know is that turnover ... a gambler losing and being replaced .... is an evolutionary event. So the rate of evolution is the rate of loss and replacement of gamblers.

In a world in which all gamblers play 'perfectly' ... they discard, bet, fold, etc. on the basis of a perfect probability model for the game they are playing ... the rate of evolution will be set by house rules of the game that is being played, including and especially betting minima (like the ante) and the betting maximum. Otherwise everything will be random. But in this model, the gamblers don't know much about playing the game, but are allowed to learn over time and get better. If you want, we can allow the gamblers to learn from each other and we can allow them to move from table to nearby table. The details don't matter. All we need to know is that clones have the knowledge of how to play the game of the individual from which they are cloned, and that over time, knowledge can improve by innovation that spreads from its source across the room.

The final element is this: Every now and then the game changes. Prior knowledge of how to play the game is then not fully accurate (though it may still be useful) because the new game has different rules ... maybe even different cards or dealing practices.

There are all kinds of things that are interesting about evolution that are not in his model, but you an add them on your own. All I want now is randomness and a certain amount of adaptation.

If all of the above is true, the rate of evolution averaged out evenly across a very small number of tables in this room and over a few hours of play is even and consistent. Everybody is acting the same way and doing essentially the same thing and change happens at a constant rate plus or minus a small amount of variation that over time mostly goes away. However, on the surface ... looking at the behavior of the gamblers .... one will see a lot of not much change punctuated by the occasional moment of novelty in the form of change in play behavior or reactions to changes in rules. One will see most gamblers disappearing (when they run out of money) randomly, but occasionally a novel way of playing that is better will emerge and a wave of more rapid replacement of the gamblers that don't pick up on this change occurring. But, that is not faster evolution. The hands are being dealt, the pots built up, the cards shown, etc. etc. at the same rate. Gamblers therefore lose their last dollar and are replaced at the same rate. The rate at which the 'genes' are doing things is always the same. What is happening here is that the phenotype (which here is card game playing strategy interacting with the game rules) staying mostly the same but occasionally changing quickly. So one cold say that the rate of adaptive evolution is in fits and starts (like punctuated equilibira), but the vast majority of evolution ... randomly being the guy at the table with the least money then losing it all ... is unrelated to adaptation and proceeds at the same pace all the time.

Now imagine taking a sample of two groups of tables from different sides of the room. One can measure the difference between them as a kind of evolutionary phylogenetic measure similar to a genetic distance. Most of the time, any two such pairs of samples will have a difference that is related to distance across this very large room (space and time are the same thing in this model, really). Physically close sets of tables will be more similar than distant onces in proportion to the distance.

But there will be glitches. Some of the glitches will be because of random variation. But if you always take samples from pretty far away this will mostly even out. Then there will be glitches because adaptive change has happened (and is likely still underway) and that temporarily un-evens the changes so you get one group changing more rapidly phenotypically than others. But again, that evens out eventually.

There is something that can cause a change in rate that is real and won't even out. This would be a change in house rules for one part of the room. Let's say that in one part of the room it is necessary to ante ten times as much as in another part of the room. In this particularly expensive area, the minimum bet is much higher than in other parts of the room, and the maximum bet is very high. Under these conditions, the probability of a gambler acting rationally losing all his/her money at once is elevated, and thus turnover will be elevated. (Note: The long term probabilities will not change ... that is based on the game rules. But this change in betting rules increases the rate at which a given individual will pass the 'vanishing line' of having no money.)

If one part of the room ... or one set of gamblers in one part of a room, or those playing a particular novel version of the game that was imposed on one part of the room ... has this difference from the rest of the room, then the rate of evolutionary change will be different This does not mean that the rate of adaptation will be different, just the rate of background change in what gamblers there are an how rapidly they are replaced. Thus, as rules and adaptations to rules spreads, the rate of evolution will actually change over time as well.

This would be a true change in evolutionary rate that actually may have nothing to do with 'phenotype.'

Other kinds of changes could happen as well. For instance, changing the game as mentioned above could change the rate of evolution. This would be akin, in real life, to some difference emerging in DNA mutation rate or repair mechanisms. This model could go all sorts of places (mainly places we don't want to go). But the main point is this: The rate of evolution in this model is not closely related to adaptive change most of the time (but it can be) and the rate of evolution in this model is not observable from surface changes quite a bit of the time.

Some, but not all, of the statements made above can be addressed more effectively with this model in mind. Let's try:

The idea that evolution is very slow, takes millions of years (for speciation), etc. is not really addressed by this model directly, but is indirectly. Indirectly what we see here is the fallacy of equating evolutionary change with speciation. Speciation is important and it is a kind of evolutionary event, but if you want to talk about the rate of speciation, then call it the "rate of speciation" not the "rate of evolution." They are not the same thing.

Not addressed in this model is the simple fact that speciation does NOT take millions of years. How could that be? The average mammal species, for instance, exists mostly unchanged for a couple million years before going extinct. In a time frame where change can be (and this is true) described in terms of punctuate equilibria, how can species that exist for about two million years on average evolve from one to the other over a course of 'millions and millions' of years. No. Species emerge over periods of thousands of years.

The statement that "Evolution has stopped for humans" is totally dumb for so many reasons that I will not go into here. We can address that later when I review Henry Harpending's new book (coming out soon) on recent human evolution. But the present model (the gamblers) shows us how evolution does not really ever stop even if on the surface nothing seems to be happening for a while. (With humans, of course, there is lots of stuff happening evolutionarily anyway.)

The relationship between evolution and geological time scales is interesting. I don't have a clue what a "geological time scale" is in this sense. I think what people mean is that there is a minimal unit of time across which we see very little or zero resolution. This is, of course not how geological data are really manifest. There is a range of scales, and some things are reliably represented at certain scales, and all scales are represented somewhere by something. The bone of an animal was formed during the life time of one animal, perhaps years or decades. The varve in a deep lake deposit may have been formed in a week or a few weeks, and the varve ten varves above was formed during a similar interval ten years later. Just because things happened a long time ago does not mean that they took a long time to happen. What is true is that the frequency of data points that represent short intervals that can be inter calibrated in geological settings is always low, and probably gets lower as you go back in time over the last half billion years or so, probably not as a percentage of the record, but in absolute occurrence.

The idea that evolution = natural selection is the biggest bugaboo of them all. It does not. Adaptive change is one kind of change, but probably a minority of the overall change that occurs at the genetic level (but maybe not at the phenotypic level).