"The Science Book!"

[sfs]

Natural selection can't in itself keep things from going downhill. NS only removes the most unfit.

Which is to say, NS removes precisely those individuals that are more downhill than the current population. If all of the individuals who are farther downhill are removed by NS, what makes things go downhill?

 

[gluadys]

Natural selection can't in itself keep things from going downhill. NS only removes the most unfit. Mount Improbable completely ignore this fact.

Natural selection can't prevent individuals being born who are less fit than the current species norm. But it can and does prevent them from carrying the whole species with them. Removing the unfit is what keeps things from going downhill.


When you read the book, you will see this explained.

 

[Smidlee]

Natural selection can't prevent individuals being born who are less fit than the current species norm. But it can and does prevent them from carrying the whole species with them. Removing the unfit is what keeps things from going downhill.


When you read the book, you will see this explained.

Like Genetic Entropy? NS selects as a whole and can't detect small genetic chances unless in life or death situation. Bad mutations comes along for the ride of the beneficial ones. NS select out the worst mutations is not the same as it can filter out all bad mutations.

 

[sfs]

Like Genetic Entropy?

I'm not sure what it is that's supposed to be like "genetic entropy". If you mean Sanford's idea of genetic decay that can't be prevented by natural selection, no, that doesn't cause real populations to decay in fitness, except under peculiar conditions.

 

[Smidlee]

I'm not sure what it is that's supposed to be like "genetic entropy". If you mean Sanford's idea of genetic decay that can't be prevented by natural selection, no, that doesn't cause real populations to decay in fitness, except under peculiar conditions.

Of course Sanford is not the first nor the last who have serious doubts about how useful NS really is. So far it has no more power than the artificial (man's) selection.
(If Sanford is correct or not doesn't do away from the fact NS would still select the same way if all living creatures as a whole are gradually going downhill)

 

[sfs]

Of course Sanford is not the first nor the last who have serious doubts about how useful NS really is.

True, but until one of them can present a scientific case supporting their doubts, that fact doesn't really signify much.

So far it has no more power than the artificial (man's) selection.

It usually has less, in fact. So what? That power is more than adequate to do what you said it couldn't do.

 

[Smidlee]

True, but until one of them can present a scientific case supporting their doubts, that fact doesn't really signify much.

It usually has less, in fact. So what? That power is more than adequate to do what you said it couldn't do.

We do have examples of genetic entropy from man's selection. It has been known for years that pure breed dogs are picking up more genetic defects by each generation.

 

[Papias]

Smidlee wrote:

NS selects as a whole and can't detect small genetic chances unless in life or death situation. Bad mutations comes along for the ride of the beneficial ones. NS select out the worst mutations is not the same as it can filter out all bad mutations.

Simply false. NS selects against all harmful mutations, to the degree that they are harmful. It's obvious that a lethal mutation is strongly selected against, true, but a non-lethal, harmful mutation that reduces the odds of reproductive success by 50% (instead of the 100% of a lethal one) is still selected against - 50% as strongly. OK, what about even smaller effects - say one that gives just 20% more or less offspring, on average?


For example - I run industrial experiments, which often have a dozen variables at once (hence called "multivariate"), in experimental sets using hundreds of experiments. Because all effects affect each result, all variables are effectively being tested in each single experiment. The upshot is that we can see which variable have even very small effects - say, a 9% effect, or less - and we can test them all at once, in just a few weeks.

Now, the same thing is going on with natural selection, except that instead of hundreds of experiments, you have literally many thousands or millions (each individual in the population), and over time, even a few thousand years, or millions, you have literally billions or trillions. I can get resolution to see single digit sized effects in just a few hundred experiments, so it is hardly surprising to expect that NS could resolve effects that only increase or decrease reproductive success by 0.001% or less, when it is running trillions of experiments.

The creationists who told smidlee that they "aren't selected against" are counting on the fact that people often don't understand how to deal with small probabilities over a large population (multiplying a very small number by a very large number). So they use the "genetic entropy" PRATT, and people believe them. Most people aren't running multivariate experiments, so they don't realize that you are testing all the mutations at once - you don't have to "wait" until the, say, ear mutation is resolved before testing the "eye" mutation, and so on.

In fact, computer models are done on this all the time, and they confirm the same thing - that even a very small detriment, say, a reproductive chance reduced from 20% to 19.9%, is weeded out by natural selection, even if artificially boosted to a lot of the population to start with.

In a day to day situation, it's really cool to see how powerful multivariate experimentation is.

Papias

P. S. the dog example doesn't argue against this because the reproduction of dogs is controlled by humans. For example - if a detrimental mutation - say, one that slows the dogs running by 20% - is present, it won't effect reproductive success at all if the breeder is breeding on the basis of how fluffy the fur it. Detrimental mutations are expected to accumulate then the selection is based on a narrow set of traits set by the breeder instead of the harsher and much more general (sensitive to all factors) selection of NS.

 

[gluadys]

Like Genetic Entropy? NS selects as a whole and can't detect small genetic chances unless in life or death situation. Bad mutations comes along for the ride of the beneficial ones. NS select out the worst mutations is not the same as it can filter out all bad mutations.

That's because NS doesn't select mutations. It selects for fitness. And it doesn't select for optimum fitness, but for the fittest among available options.

Sure bad mutations ride along with beneficial ones, but if the combination is the most fit among the variations in the population, then it will receive positive selection. And there will be no "going downhill" to a less fit condition as the norm for the species.

btw, it is not true that NS is only operative in life or death situations.

We do have examples of genetic entropy from man's selection. It has been known for years that pure breed dogs are picking up more genetic defects by each generation.

From human selection, yes. Because human breeders don't select for fitness.
They select for show and for human (not canine) benefit.

 

[Smidlee]

That's because NS doesn't select mutations. It selects for fitness. And it doesn't select for optimum fitness, but for the fittest among available options.

Sure bad mutations ride along with beneficial ones, but if the combination is the most fit among the variations in the population, then it will receive positive selection. And there will be no "going downhill" to a less fit condition as the norm for the species.

btw, it is not true that NS is only operative in life or death situations.



From human selection, yes. Because human breeders don't select for fitness.
They select for show and for human (not canine) benefit.

Humans still doesn't want a defected dog yet these defected are increasing in spite. The dog's characteristics are being selected for.

In fact, computer models are done on this all the time, and they confirm the same thing - that even a very small detriment, say, a reproductive chance reduced from 20% to 19.9%, is weeded out by natural selection, even if artificially boosted to a lot of the population to start with.

I have serious doubts NS can detect a small reduction from 20% to 19.9% when you add in plain old luck. There is a lot of luck involved in nature in who survives which put a limit on NS can select. Some have mention if survival of the luckiest is a bigger factor than survival of the fittest. Now this is an advantage in man's selection as we can remove luck from being a major factor. Yet even though we can remove luck we haven't so far remove "genetic entropy" from our selection even if we select on different principles .

 

[samaus12345]

Selection can only act on whats available. The myth is 'neofunctionalization' is more than what it is, a myth. Only when NEW functional proteins arise that influence the phenotype and make it more/less fit can selection act on it. Again Richard Dawkins wasnt exactly 'stumped' by that question. He is a zoologist. He knows biology. He KNOWS such a mutation has never been observed (i dont know if hes so self deluded that he has convinced himself that its possible or if he knows it is impossible, i tend toward the second one because hes not a complete monkey, he has done a good job deceiving people into believing something he knows, at the very least, has never been observed).

 

[sfs]

Humans still doesn't want a defected dog yet these defected are increasing in spite. The dog's characteristics are being selected for.

Strong selection for one trait can easily drive linked deleterious mutations to higher frequency. Extreme inbreeding can also lead to an increase in frequency for a deleterious mutation. Humans impose both of these conditions on dog breeds, so it's hardly surprising that deleterious mutations have appeared in dog breeds. All of this is well understood by evolutionary biology, and none of it undercuts evolution at all.

I have serious doubts NS can detect a small reduction from 20% to 19.9% when you add in plain old luck. There is a lot of luck involved in nature in who survives which put a limit on NS can select. Some have mention if survival of the luckiest is a bigger factor than survival of the fittest.

Yes, NS does indeed have to compete with randomness. Evolutionary biologists have been aware of this quite obvious fact for many decades, and have long since developed (and tested experimentally) a theoretical framework for understanding it quantitatively. How small an advantage or disadvantage NS can detect depends on the size of the population; on large populations, NS can drive mutations that have much less than a 1% advantage.

 

[sfs]

We do have examples of genetic entropy from man's selection. It has been known for years that pure breed dogs are picking up more genetic defects by each generation.

This is not what Sanford means by "genetic entropy". He's wrong too, but at least he's wrong in an interesting way.

 

[samaus12345]

"Hes wrong". Data?

 

[Papias]

Smidlee - sorry for the delay, I took the family camping.

Smidlee wrote:

Humans still doesn't want a defected dog yet these defected are increasing in spite. The dog's characteristics are being selected for.

Sure. That's because the deletrious effects may not show up until well into the dogs reproductive life, after they've had one or more litters, thus, the deletrious genes are still passed on. If dog A has a harmful mutation that kills the dog upon reaching 60% of the way through it's reproductive lifespan, then that harmful gene is selected against more weakly than a gene that causes dog B to be 50% less likely to be re-bred. And, since genes are always in a competitive situation, that means that dog A will be selected FOR by humans, over dog B, thus increasing the accumulation of harmful genes.

Originally Posted by Papias

In fact, computer models are done on this all the time, and they confirm the same thing - that even a very small detriment, say, a reproductive chance reduced from 20% to 19.9%, is weeded out by natural selection, even if artificially boosted to a lot of the population to start with.

I have serious doubts NS can detect a small reduction from 20% to 19.9% when you add in plain old luck. There is a lot of luck involved in nature in who survives which put a limit on NS can select.

Plain old luck is included, and over a large sample size, is irrelevant. This is due to "regression to the mean". Sure, in one individual, luck is most important. But since there is no such thing as "being lucky", there will be an approximately equal number benefitted and hurt by chance. On average, some of these will have good genes and some bad, and those effects will be added, such that over a large sample size, even the tiny effects of genes will be amplified.

For instance, If I flipped a coin 10 times, I may very well end up with 80% heads, even with a fair coin. But if I flipped a coin 100,000 times, there is no way I'd get 80% heads, and a result outside of the range of 49.5 to 50.5% would be very unlikely.

As people, who deal with day to day things and of whom very few of us actually flip a coin 100,000 times, this reality is easy to be unaware of. However, it's real, and it's why even small genetic benefits win out over large sample sizes, regardless of chance.

Some have mention if survival of the luckiest is a bigger factor than survival of the fittest.

As explained above, it is easily the biggest factor for an individual, but is averaged out over a population.


Make sense?

Papias

 

[Smidlee]

As explained above, it is easily the biggest factor for an individual, but is averaged out over a population.


Make sense?

Papias

A beneficial gene must first appear to an individual but it becomes fits in a population. Let say an individual got lucky did have the next evolving gene but got killed in a car accident at age 8 yet their sister was unlucky and had a bad mutation yet survived and had 8 children. natural selection couldn't pick the fittest because of bad luck. Natural selection has to wait until someone happen to get lucky and survives to reproduce. Then there the problem of marrying someone who probably carries bad mutations. There offspring will has a mix of the beneficial mutation along with the bad ones.
You need large population to get lucky enough to find a benefical mutation (small step that Darwinism requires) but then you need a very small population in order to have any chance of the mutation becoming fixed. Evolution has a serious problem with sex.

 

[gluadys]

A beneficial gene must first appear to an individual but it becomes fits in a population. Let say an individual got lucky did have the next evolving gene but got killed in a car accident at age 8 yet their sister was unlucky and had a bad mutation yet survived and had 8 children. natural selection couldn't pick the fittest because of bad luck.

Unlike you, natural selection doesn't measure "fittest" against a non-existent ideal. It measures "fittest" against what actually is. If someone dies before they can reproduce, any gene they have doesn't count when measuring fitness. Fitness pertains only to the actually reproducing population.

It also takes more than having children. One has to have children who survive. The mother of eight children might see no grandchildren because her children did not live to reproduce.


So you can't predict the fate of a mutation (good or bad) in a population based on its fate in an individual. You have to look at all the data for the whole population and over several generations.

Natural selection has to wait until someone happen to get lucky and survives to reproduce. Then there the problem of marrying someone who probably carries bad mutations. There offspring will has a mix of the beneficial mutation along with the bad ones.

And natural selection handles all these problems easily because it selects fitness by genomes not by individual genes.


Do you know how to run an algorithm for a selection factor? If you do, set up some programs and run them to see for yourself.

You need large population to get lucky enough to find a benefical mutation (small step that Darwinism requires) but then you need a very small population in order to have any chance of the mutation becoming fixed. Evolution has a serious problem with sex.

Not really. There is usually enough variation, even in a small population, that some alleles will provide a benefit that other alleles don't. And those will be favoured by natural selection. Remember natural selection depends on the presence of variation, new or old, not just on the occurrence of new mutations. So any existing variation can be subject to selection.

The one scenario in which there may be too little variation to choose from would be when a small population also consists entirely of closely related individuals.

 

[Papias]

Yep, Smidlee, good points. Let's look at them.

First, it seems that we agree that once a beneficial mutation is in a good number of individuals (say, several dozen), that it's average benefit will cause it to spread to the whole population.

The topic now is that earlier stage.

Smidlee wrote:

A beneficial gene must first appear to an individual but it becomes fits in a population. Let say an individual got lucky did have the next evolving gene but got killed in a car accident at age 8 yet their sister was unlucky and had a bad mutation yet survived and had 8 children. natural selection couldn't pick the fittest because of bad luck.

Right. For beneficial mutations with very small benefits (like my 19.9 vs 20% benefit to survival example given previously), individual chance will be more important when looking at the first mutational generation. At the same time, remember that in every generation, a sizeable proportion do in fact survive, and so the mutant as at least those odds (and in fact a touch better odds). So as numerous beneficial mutations appear, many are indeed removed by chance, but also by chance, some will survive, and thus their odds are even better the next round (because they will be in a number of offspring). So yes, it is undoubtedly true that literally billions of mutants with beneficial mutations died due to chance - but at the same time, chance isn't "unfairly tilted" against them either, so over many times that benefical mutations occur, the odds (regression to the mean) dictate that natural selection's influence will still be relevant.

Natural selection has to wait until someone happen to get lucky and survives to reproduce.

Yes, a beneficial mutation has to make to to the next generation, and many no doubt fail, yet their odds are no worse (indeed slightly better) than anyone else's so over many appearances of beneficial mutations, they will, on average, increase in number. Also importantly, don't fall into the trap of thinking that they have to be sequential. You don't have to "wait" for mutation "A" before "waiting" for mutation "B". Genomes are combinatoric, and multiple beneficial mutations are spreading in the population at any given time. For instance, look at skin color. At least a dozen genes affect skin color, so a light skinned group moving into a sunny area will have an overall selective pressure that would favor "darker skinned" mutations in any of those 12 places. A dark skin mutation in locus 8 would be selected for, while three families over a dark skin mutation in locus 2 is also selected for.

You need large population to get lucky enough to find a benefical mutation (small step that Darwinism requires) but then you need a very small population in order to have any chance of the mutation becoming fixed.

No, exponential growth shows that a beneficial mutation can become common in a large population as well, and even in a small population, there are many thousands of births over a number of years.

Then there the problem of marrying someone who probably carries bad mutations. There offspring will has a mix of the beneficial mutation along with the bad ones. ...Evolution has a serious problem with sex.

That is known as the "blending problem". For example - say you had brown rabbits who moved into a snowy region, and one day there was a beneficial mutation that gave one rabbit white fur. That rabbit would be very successful, and say it had 10 kids, which, having a brown mother, would be tan. Those would be less successful, and have kids that (becuase they mated with regular rabbits), be 3/4 dark - not nearly as good anymore. Within a few generations, the beneficial effect is lost. Big problem for evolution, right?

Mendel's work showed that the real world doesn't work like this, but that genes are discrete, so the kids aren't all uniform, and the traits aren't completely blended (as, say, mixing paint cans would be). So in the first generation of 10 kids, some would be as dark as their mother, and some much lighter. NS would mean that the lighter kids would be selected for, and in the next generation (including some interbreeding), some would be white, completely dark and maybe inbetween. NS would again benefit the whitest rabbits, and this would accelerate as the genes for white fur spread in the population.

The "blending hypothesis" was seen as a serious challenge to NS in the late 1800's. However, when actual inheritance was figured out, it was seen that the "blending hypothesis" is incorrect, and thus poses no problem to evolution. In your example, the random assortment in genes mean that some children will have some of the harmful mutations, and some wont', and some kids will have the beneficial ones and some won't - thus NS can work on the various genes, even though all the kids had the same parents. You can see this in your own kids - they aren't clones of each other. I sure can see that in mine.


Blending inheritance - Wikipedia, the free encyclopedia

Gluadys, thanks for also explaining many of these points - I'm sure that some will find gluadys' explainations easier to understand.

Papias