There are seven living species in the genus Equus: two horses, two asses, and three zebras. They really act and look pretty much alike, aside from the color scheme. So, they obviously have about the same genetic information. But they don't have that information spread across the same number of chromosomes. One of the zebras has only 32, and Przewalski's wild horse has 66. And, all of these species follow the breeding structure of a harem, with kin breeding. That is, there are family groups, that breed only among themselves for generations in a row.
So, it seems very likely that these species were created by chromosomal speciation. Here's how it works.
Suppose that a single mutation happened when a single stallion was conceived. Suppose that that mutation was not a change to a gene, or the loss of a gene, or anything like that. Suppose that it was instead a restructuring of where things were on the stallion's chromosomes. Perhaps it was an inversion mutation, or translocation mutation. Perhaps the translocation moved stuff to a whole new chromosome, thus changing the number of chromosomes.
Now, this would pose a problem for most males. There aren't any females (yet) who have that mutation. This means that the stallion is most likely facing a really serious fertility problem. However, he has the huge advantage of owning a harem. It doesn't really matter that much if he has less than the usual number of offspring. As long as there is at least one female offspring, the mutation has passed its biggest hurdle.
The next problem is that his offspring are hybrid. That is, they acquired normal genetics from their mother, and mutated genetics from the stallion. With a herd or flock species, this could easily put the offspring at some sort of disadvantage, compared to the "normal" animals. However, since a kin group is closed to outsiders, the hybrids have no competition. That increases the chances that a few will live.
This brings us to what happens when the stallion's offspring try to mate. By Mendel's rules, each grandchild has a one-half chance of getting the mutation from its mother, and has a one-half chance of getting the mutation from its father. So, there is a one-quarter chance of the grandchild getting two copies of the mutation. Again, fertility may be low. But assume for the moment that one such male is born, and one such female is born.
The mutation is now all set. There is a male and a female which have the mutation on both sides. They can breed without any fertility problem at all, and can start trying to spread their kind across the earth. There is now a new species. Their interfertility with the old species would be similar to what we see between horses and donkeys, except better, since horses and donkeys appear to be several mutations apart.
Notice how many assumptions we've made. Given the social structure of horses, it really only comes down to supposing a mutation which is just big enough, and also just small enough. (If it's too big, there won't be children. If the mutation isn't big enough, the mutant's children will breed with normal horses, and the mutation just gets lost in the crowd and forgotten.) We know that that sort of mutation can happen, and it most likely wouldn't affect the viability of the resulting species. And, the mutation would leave chromosomes just like the ones we find in those animals today.
I don't have measured numbers, so let's make some up, to illustrate the big picture. A birth defect occurs per 100 or so births. Let's say that our mutation happens much less: let's say once per million births. Suppose the mutation has one chance in 100 of passing the fertility hurdle. Then, it has once chance in 100 that the new species will thrive and spread. Geological dating shows that species typically last for a few million years. Assume that one generation is one year. So, if we want one new speciation per million years, simple multiplication says we need a population with 10,000 males. That's quite consistent with the number of caribou alive today, or the number of wildebeeste alive today.
The scenario assumes some special conditions. So, this is probably not the commonest mechanism.