Natural Selection

The tendency for sub-populations within a larger population to increase their proportional representation within that population in relation to their relative reproductive fitness.


(moved from DefinitionOfLife)

Discussion on the importance of NaturalSelection...

Here's one: Something is alive iff, within an arbitrary range of time, it replicates such that the replicant is also capable of replicating, and the replicant is different from the original, and those differences are both inheritable and also affect the replicant's ability to reproduce when compared to the original.

In other words, something is alive iff it is under the domain of NaturalSelection. -- RobHarwood

Not so simple. Fire fits this (and many other) definitions of "life". -- AndyPierce

The differences in fire do not affect the fire's ability to reproduce, and also those differences are not inheritable, they are random.

I humbly submit that fire can be a process in a LifeCycle?. Australian flora, for example, cannot reproduce without regular fires to crack their seedpods. Also, you seem to be suggesting that domesticated species, whose genetic diversity is fast approaching zero, are in imminent danger of extinction, which they aren't unless we are.

Fire: No argument there. In order to reproduce, organisms must interact with their environment. So sure, the environment is part of the LifeCycle?. And fire is part of the environment. But I think it would harm the usefulness of the definition of life if fire was included as 'alive'.

Very few people think fire is alive. The problem is that a naive definition of life usually includes it...that's why it's a good test case. In high school biology we took the bigoted route out of this and said "life must be made of cells"...

I agree. Simple crystals, and automobiles are other good test cases.

Domestication and genetic diversity: Non-zero genetic diversity is absolutely a requirement by the above definition. However, for all practical purposes, zero genetic diversity in natural life rarely occurs on anything more than a trivial scale. There is still a lot of variation in domesticated animals, as well as cloned organisms such as asexual bacteria. Mutation is ubiquitous in natural life.

This is a useful thing, though, only if you have a working population. A single bacterium has no genetic diversity; it's offspring will usually end up identical to it (assuming it survives that long). There's no non-arbitrary way to group bacteria into populations, since the difference between mutated daughters and different divisions of bacteria is only one of degrees (no interbreeding). So how do you deal with one?

The definition above requires at least one act of reproduction. You cannot determine if something is alive if you aren't able to observe it reproducing (well, I guess you could make educated guesses in some circumstances...). Anyway, the 'range of time' bit is important. It is within this arbitrary time that the replication is observed. So now you have a copy. You compare the copies. If, as you propose, the copy is exactly identical in all fitness determining characteristics, then you cannot determine that the object is alive. Consider this thought experiment to see why this is important:

You observe an object replicating without modification. Each copy is identical to the next. Soon the copies fill up whatever container you were observing them in. But since each copy is identical, there are no differences in fitness, so NaturalSelection has nothing to work on. There is no evolution. Seeding a supersaturated solution is an example of such a situation. Crystals form which cause other crystals to form, etc., but none of the replications exhibit evolution.

Until genetic variation occurs, you are essentially talking about the same object (for the purposes of discussing EvolutionByNaturalSelection).

I agree that in order to proclaim something as alive, you have to watch at least one reproductive cycle. However, most of the time when bacteria split, the daughters are genetically identical to the parent. You could watch several generations and still not witness any mutations. Asexual eukaryotes are even worse; mutations almost never happen.

Almost never, as in, all the time? (see below) - and yes (see below)

Note also that two physically separated bacteria are not the same individual, genetically identical or no. They might be practically identical as far as evolution is concerned, but evolution isn't everything (see below). Are identical twins actually just one object?

The qualification 'for the purposes of discussing EvolutionByNaturalSelection' is important, of course (otherwise I wouldn't have included it). We're discussing the definition of life, not the definition of individual.

Life isn't some sort of play-dough, it is a quality that living things have. Since we are making reproduction part of the criteria, we have to be able to decide when two things are separate. And the qualification is unfair and irrelevant. We're discussing the definition of life, not the definition of EvolvingUnit?, and you're wrong to simply assume they're the same.

The idea that NaturalSelection is the central characteristic of life has, well, issues. Any replicator which shows potential for change will undergo NaturalSelection, simply because stuff happens. Making that the essence of life would include weird things like memes or song remixes that aren't particularly alive. On the other hand, NaturalSelection has most living things try their hardest to escape mutations. Most mutations, after all, are neutral or bad. Eukaryotes have a brilliant self-correction system and the asexual ones only evolve when it slips up. Are we saying that the essence of life is that it doesn't work properly?

Properly according to whom?? - You know what I mean. Change is not inherent in the way something like a Euglena works. It happens because even after triple-checking, you still make a typo every once in a while. I feel justified in saying the enzymes etc constitute are meant for perfect copying because evolution often selects for perfect copiers, most mutations being bad and Euglena being pretty much good enough already.

Evolution will never select for perfect copiers in a dynamic environment. In fact, the more dynamic the environment, the further away from zero the mutation will be. This is not 'improper' as you suggest (who's to judge 'proper' anyway?), since it can be shown that in environments of differing dynamism, the most fit will be those species that tune their mutation rate to match their environment.

Would you please AnswerMe when I ask: where can I find more information on the differences in mutation rates between species? I've also heard that specific regions ("hypervariable regions") of a single organism's DNA have higher mutation rates than the rest, but I don't know where to look for more details. -- DavidCary

[Try http://www.pnas.org/cgi/content/full/99/2/803 (found using Google).]

Even in a static environment, such as a toy problem with GeneticAlgorithms, where evolution will tend to converge on one solution, I'm perfectly comfortable saying that the system died through convergence. Evolution is no longer occurring, so in my mind, and by the definition above, there is no longer life.

It wouldn't hurt to sometimes consider biology. Very few systems are dynamic enough to require a mutation every time a protozoan reproduces. Most environments are actually quite stable on that scale, e.g. fresh water. Fluctuations usually can be shielded against, rather than adapted to. And as a result, mutations are bad more often the good.

The result is that protozoa have developed elaborate biologies to prevent mutations from occurring. These are indeed selected for, since a better copying system means more accurate (and non-defective) offspring. Look at the rate of mutation in any asexual protist; they've come very, very close to convergence. And yet these are some of the most ubiquitous life on the planet.

It's like a SearchAlgorithm?. The algorithm is 'running' if it is approaching the solution. If it finds the solution, it ceases 'running'. If the solution keeps changing, then the algorithm (if it's robust like NaturalSelection) will continue 'running'. Algorithm = subject under consideration. Running = alive. Changing solution = dynamic environment. I have coined the term NaturalSearch to encompass this concept.

The problem is that running = alive in this sense is nothing like what life means in ordinary biology (cf RK's comments). If you had several generations of no change in a bacterium population, it would be dead, despite them continuing to metabolize and reproduce? If a cosmic ray struck one and changed a base, would it have been briefly resurrected? You are giving life a completely new definition, incompatible with its previous one.


The things common-sense humans classify as "alive" has nothing to do with things that are "under the domain of NaturalSelection". Humans classify many things as "alive" that are not "under the domain of NaturalSelection" because they cannot even replicate - for example, sterile hybrids such as mules are alive - while they classify many other things as "not alive" that do seem to exhibit NaturalSelection - for example, the sub-population of Wal-Mart stores within the larger population of supermarkets, or the population of "fish-hooks with the barb on the outside" within the larger population of fish-hooks, or variations on popular folk-songs. -- DavidCary

What if some "object" has more than one life. Humans could have two distinguishable lives, which both fall under the definition of NaturalSelection, Evolution or Life, so that both lives must die before the human could be classified as dead (source: http://www.su.se/~chan0416/life/).
See NaturalSearch, DefinitionOfNatural, DefinitionOfLife, EvolutionByNaturalSelection
CategoryEvolution

EditText of this page (last edited February 20, 2005) or FindPage with title or text search