Carl Zimmer in The Loom:
When biologists think about the evolution of life, they think about climbing mountains.
To understand their alpine frame of mind, imagine a biologist studying the fish in a lake. Each fish may be born big or small. Fish born at certain sizes may be more likely to survive and reproduce than others. Each fish may be aggressive or shy. Again, their aggressiveness may determine their odds of having babies.
To picture all of this, it’s very helpful to imagine a landscape. Each point on that landscape is a different combination of aggression and body size. They’re like the longitude and latitude on a map. Each combination leads to a particular level of reproductive success. Picture that success as the elevation of that point on the landscape. The more success, the higher the altitude.
The biology of those fish can give the landscape a topography. Perhaps it produces a single mountain. The peak is the combination of weight and aggression that produces the most possible babies. The landscape drops off in all directions, to combinations that make it more likely the fish will die, or fail to reproduce.
The actual fish in the actual lake might turn out to be clustered on one of the mountain’s flanks. The fish closer to the peak have more babies than the others further downhill. As a result, they’ll pass down more copies of their genes to the next generation. And that means that the population will climb up towards the peak. If a new mutation arises, natural selection will favor it if it helps the fish climb further. Eventually, the fish may reach the mountaintop. Once they plant their flag on the peak, they’ll be stuck. Natural selection won’t be able to nudge them off.
Now imagine that there are two peaks, not one. The fish sit on one mountaintop, while the second peak towers over them in the distance. They can’t get to that second peak, though, because natural selection can only nudge them uphill. They are stuck with a mediocre body.