I know the OP says the question has been resolved to his satisfaction by David B’s “species do split” explanation (well-written and enlightening as always), but let me add something for the benefit of anyone else reading. I’ll try to explain the OP another way, by using a simplistic thought experiment. (I used a similar example in a previous thread maybe a year ago, about punctuated equilibrium.)
Imagine a valley. High ridges on both sides isolate the valley from neighboring features. The valley’s sole plant life is a type of grass that grows to a certain height.
Living in this grass is a mouselike creature, an insectivore. The mouse’s main predator is a larger animal, like a weasel. The mouse is well-adapted to living in the grass; through thousands of years of adaptation, it grows to exactly the right height that it is hidden in the grass, but it can stand up and look around to check for danger. Its legs are the right length to balance speed and agility in the grass.
Random mutations continuously occur, as they always do. Some mice are born taller, which means they can’t hide in the grass as effectively and get eaten more often, which means they are at a disadvantage for survival. Some mice are born shorter, which means they can’t stand up and look over the grass as effectively, so they can’t identify predators as well; same result. Mutations occur, but the alternative traits are inferior to the existing set of traits, so they never take hold, and the species maintains status quo.
Now imagine a volcanic eruption nearby. The falling ash changes the chemical composition of the soil slightly, and simultaneously the winds that have been temporarily changed by the eruption bring a new type of grass seed into the valley. The old grass doesn’t like the new soil chemistry, but the new grass does, so the new grass replaces the old grass in short order.
This new grass isn’t as tall as the old grass. Immediately, the existing mouse “design” doesn’t work as well, because they’re too big and can’t hide in the grass. (I use the word “design” improperly, I know; this is a simplified example. Just bear in mind it is not intended to convey any sort of conscious or deliberate modification of the mouse’s body structure. I just can’t think of a better word at the moment.) The random mutations continue, though, and the same shorter mouse that previously was at a competitive disadvantage is now suddenly a superior “design,” because it’s the best height for the now-shorter grass. Hence, the “old-style” mouse dies out quickly, and if the “new-style” mouse can reproduce fast enough, it replaces the old mouse in the ecosystem. At this point, they may or may not have speciated; this might be a simple characteristic.
Now change the scenario slightly. There’s a river at the bottom of the valley that separates the two sides. The same grass and the same mouse occupy both sides, because even though they’re separated they have no reason to change.
The volcanic eruption, due to wind patterns, brings ash and grass seed to only one side of the valley. The changes described above occur only on that side. The old grass and the old mouse still survive on the non-changed side. Again, at this point, speciation may or may not have occurred, but the slight difference in the two types of mice is at least the first step toward it. (Speciation, of course, has been extensively covered in other threads, and I’ll leave it alone.)
I know this seems like a silly example, and in many ways it is. A real ecosystem is far more complex, with many more interacting elements. However, the isolated valley is well-known in evolutionary biology; read Privileged Hands: A Remarkable Scientific Life by Geerat Vermeij. He’s one of the world’s pre-eminent experts in mollusk evolution, specifically snails. In the book, he describes his work in the valleys of New Zealand (among other things), where he’s found that the various species of land snail are almost invariably unique from valley to valley, and often from one side of each valley to the other, depending on sun exposure, rain patterns, and the like.
Anyway, I hope this helps further clarify the original question.