Rosio Pavoris a blog

In Biology, fitness landscapes are a great way to visualise the relationship between phenotypes (or indirectly, obviously, genotypes) and reproductive success.
What you do is, you plot traits, or gradations of traits, or collections of traits, on axes, and put the success on an orthogonal axis. For example, if you use the X-axis and Y-axis to plot traits, you could represent success (or “fitness”) on the Z-axis, as height. Like so:

Note how it’s a hill rather than pillars. This is because 5% of an eye, for example, is still better than total blindness. This doesn’t necessarily stop it from being a rather steep hill, though.
Obviously a realistic complete fitness landscape would need a lot more dimensions than just two plus one, but that gets slightly more difficult to represent. Most fitness landscapes will tend to look at just one or two dimensions, which, while a good visual aid, can be rather tricky to do correctly.

It goes without saying that natural selection will tend to move species up the hill, and will never allow downward movement. If an individual were to be born with traits that would put him lower on the hill, he would be outcompeted by his fitter brethren and just wouldn’t pass his genes on.

Now, this leads to an apparent problem when you realise fitness landscapes don’t tend to look like the simple one in the picture. Instead, they will tend to look more like this:

The problem being, of course, that if a species gets stuck on a minor peak, it has no way of getting off it and making it to the highest one. This would lead to niche overloading, which would mean the fitness landscape would change because suddenly lack of resources would mean that there is less and less of an advantage to that combination of traits, which would lead to the peak eventually levelling off and perhaps turning into a valley, allowing the species to move on to the next peak.
This would entail a lot of individuals dying off in the process, and obviously doesn’t happen, or at the very least isn’t the whole story. So what’s the deal?

Well, you should keep in mind that a peak is an optimum; it’s the best of all possible combinations, and natural selection will make the average of the species as a whole trend to it. But a species isn’t composed of average individuals, and the distribution will often look more like this (indicated in red):

You don’t have to be the best of all possible worlds to be viable. The world is big enough to accomodate some failures for a bit. And if those failures turn out to be on the slope of a different hill, perhaps eventually their offspring will outcompete the previous optimum. Or, of course, not.