The post below on the genetics (and relaxation of constraint) of dogs has given rise to many good comments. I want to highlight one:
...Yet most feral dog populations quickly revert to a medium-sized, short-coated, yellowish dog - the so-called "Pariah Dog" that's found in so many places around the world - why don't the feral populations look more like their wolf ancestors?
There are many issues mooted below. But this comment is a good one. Why do dogs "revert" into pariah dogs instead of the Eurasian wolf? Remember what is happening here: being as "cute" as a poodle doesn't matter once you're a stray (at least unless you are adopted because of your cuteness), so the less cute dogs with skillz do better in the game of life (reproducing). And yet dogs do not quite turn back into the wolves from which they derived. Why not? One could argue that the most plausible explanation is that a) the niche of Eurasian wolves is already taken b) a new niche, that of canines which live as "urban wildlife" has opened up. But, I think there might be theoretical issues which might offer us insight into what is going on, and that insight is the Shiftling Balance (see here for a good review). The manner in which the Shifting Balance might apply is that "breeds" might be thought of as local populations, and this sort of substructure is essential for the sort of evolution which Sewall Wright envisaged. The Eurasian wolf may be assumed to be a large mixing population which is "stuck" at a fitness optimum. In contrast, dogs, artificially constrained into separate populations, allowed to "explore" a greater expanse of the adaptive landscape, can find new fitness optimums. When human selection is relaxed the nearest "wild" fitness optimum might not be that of the Eurasian wolf, a top niche predator, rather, the skillz of the domestic dog lend themselves to living as scavengers around human habitations. Please note that the Shifting Balance is not just about random genetic drift swamping the power of selection, rather, it also implies the bringing to fore gene-gene interactions as prior selection on loci are relaxed. One might conceive of the domestic dog as a new "coadapted gene complex." Of course, please do remember that the Shifting Balance is better metaphor than formal theory. But, I do think conceiving of evolutionary dynamics in this manner is a good place to start, a way to bridge the chasm between formal population genetic theory and the public discourse around evolution.