Jerry Coyne is an expert on speciation. That's why it's always informative to read his latest thoughts on the problem. In spite of what many people might believe, the main cause of reproductive isolation—the actual speciation event according to the biological species concept—is due to accident, not adaptation. It's just one more example of the importance of random genetic drift in evolution.
Here's how Coyne puts it in his latest posting [“Reinforcement” and the origin of species].
Genetic barriers aren’t thought to arise for the purpose of keeping species distinct. Rather, they are usually thought to be evolutionary accidents: geographically isolated populations diverge genetically under natural selection or other evolutionary forces like genetic drift, and that divergence leads to the evolution of genetic barriers (mate discrimination, the sterility of hybrids, ecological differences, etc.) as byproducts of evolutionary change. For example, populations could adapt to different environments (one dry, one wet, for example), leading to them becoming genetically different. When these populations meet each other again, this genetic divergence could result in hybrids that don’t develop properly because the parental genomes are sufficiently diverged that they can’t cooperate in building a single individual.I wish more people would assimilate this message. It seems to be the overwhelming consensus among the experts in speciation but the average scientist still has an adaptationist view of speciation (and of evolution in general).
Speaking of adaptationists, Coyne also likes the idea that some examples of reproductive isolation can be reinforced by natural selection. You can read about those cases on his blog.