If you've got a couple hourse to kill, listen to Stephen Stich explain why the human moral sense, such as it is, is a “hodge podge of multi-purpose kludges.” This is one of Stich's fascinating 2007 Jean-Nicod Lectures. You can follow along with the slides below the video.
Here are the slides.
One upshot, among others, is that you're not going to develop a useful normative moral theory by testing and refining your moral intuitions against cases. Another, closely related, upshot is that you can have the best scientific theory possible about the evolutionary basis of whatever sentiments and dispositions you think is central to morality, but it's not going to leave you with anything like a useful or coherent theory of the right or the good. One thing I think some ev psych fans have a hard time getting their heads around is that morality–the system of norms that regulates individual behavior and enables social coordination–is variable by “design,” and that our evolved moral capacities are largely norm-acquisition devices which must wait to be calibrated by enculturation. We're “fill-in-the-blanks slates” not blank slates. There are what you might consider “factory default” settings. (Which involves a lot of out-group slaughter, I'm afraid.) So, yes, certain configurations of moral sentiments, certain systems of norms, are more “natural” than others. But they lead to relatively terrible societies–nasty, brutish, short, etc. The norms that undergird the peaceful liberal order of impersonal, extended, massively positive-sum exchange are the result of generations of often self-conscious resistance to the “factory defaults.” Which is just to say, T.H. Huxley knew what he was talking about.