lpetrich

I think a problem with the large-mutation hypothesis of evolution is that it has a deus ex machina,. quasi-creationist quality.

This is because the mutations that get "accepted" are usually relatively small ones; large ones are usuall bad ones.

However, as understanding of molecular biology and genetics improves, it may become possible to frame testable versions of the large-mutation hypothesis, as has recently been done with the Ultrabithorax gene and arthropod abdominal legs. Brine shrimp (Artemia salina) have a version that allows abdominal legs to grow, but insects have versions that do not; putting the shrimp gene into a fruit fly causes it to attempt to grow abdominal legs.

Another case of big jumps in evolution may be the loss of forelimbs by snakes -- their Hox genes no longer distinguish the area where their forelimbs would grow, though the rear-limb area is still distinguished.

There are numerous smaller-scale cases like gene duplications, insertions, and deletions; the molecular version of gradual change is point mutations, and much more than that is known to happen at the molecular scale.

Undercurrent

This kind of relates to (at the risk of citing from Stu Kauffman) the correlation length of the fitness landscape.

The principle is, on a "sensible" landscape, two points that are near to each other, are similar to each other in fitness. For example, consider fitness as elevation on the space of longitude/latitude coordinates. If I'm on top of mount Everest, points 10cm away from me are also very high up. If I'm in the Grand Canyon, points 10cm away from me are also very low down. It's a rare condition (edge of a cliff) that a point 10cm away from a point of a given elevation is signicantly different from that elevation.

But that's not true for distances of 10000km. Knowing that point P is 10000km from Mount Everest doesn't say terribly much about point P's elevation. The best guess is that P's elevation is the average elevation of all points on the Earth -- the exact same as it would be if we didn't know that P was 10000km from mount Everest.

The moral is, that if you suck as an organism, big jumps beyond the correlation length of the fitness landscape will tend to stick around better, because one would expect them to be about average. Short jumps, on the other hand, wouldn't survive so well, because their fitness is likely close to your fitness, and you suck.

If you're really good, though, large jumps on the landscape would be selected against, because, generally, they'll only be as good as the average of the landscape, which isn't as good as you. Short jumps, however may be selected for, since they have fitness very near to yours (which is good) and maybe even a bit better.

And if you were Stu Kauffman, you's say that this applies to all systems in which there's a fitness landscape involved. For example :

• Biological Evolution -- A lot of new body plans were tried out in the Cambrian explosion when organisms in general weren't very good at being organisms, but proceeded to less drastic changes when organisms were highly optimized.
• Industry -- When airplanes were first made, a lot of new designs all over the map were tried out, but nowadays there are only a few basic, specialized designs (Passenger liner, figher jet) and most differences are just variations on a theme.
• Religion -- Back in the good old days there were tonnes of little belief systems that were kind of sucky at holding adherents. Now religious belief is concentrated into specialized faiths that are good at holding adherents (e.g. Christianity, Islam) and most variations that survive are minor (Catholic/Protestant/Mormon, Shiite/Sunni, &c.)