If you have the time (a lot of it -- should be evident from posting frequency that I'm procrastinating with a vengeance) you might be interested in this bloggingheads discussion between Don Loeb and Peter Railton about moral realism. (Watch Loeb's eyes while Railton's speaking.) I had read some of Railton's stuff at college, thanks to Alan, but it was all very meta-ethical and nonconstructive, and seemed liable to dismissal on "show me an absolute moral truth and I'll show you a contradiction" grounds. The video makes it clearer what Railton's truths are: stuff like "you should keep your promises." (See also: this Boston Review piece.) The "evidence," such as it is, comes from history and psychology. I find his position pretty unappealing, because (1) he disqualifies all forms of local feeling (patriotism, etc.) as not moral, (2) intuitions even about cheating are not quite the same everywhere -- in Tanzania, where I grew up, it's generally considered legit to cheat an out group, (3) it is not true outside of Ann Arbor that everyone believes in tolerance. Now I happen to agree with Railton's sentiments, but I'm bemused at any historical analysis implying that people have always held universalist values.
Defenses of moral truth often come down -- implicitly -- to the claim that the fact/value distinction is fuzzy, and denying moral truths is like solipsism. The difference is that we all agree, fairly strongly, on what the natural world is like, and philosophical justification is a formality; however, we do not generally agree on morality, and I have no consistent intuition about the existence of moral truths.
There's an interesting bit (ca. 30:00 to 40:00) on cognitivism: cognitivists claim that moral assertions are statements about something, whereas non-cognitivists claim that they are assertions of attitude -- e.g. yay red sox. (The BR article has a nice example of the semantic dangers here.) Non-cognitivists are obligated to explain how logic and consistency apply to moral statements, if they do. The if is something I've wondered about in the past, and it still puzzles me. There are some trivial cases of "moral reasoning" -- Killing pinnipeds is bad. A walrus is a pinniped. Therefore, killing walruses is bad. -- in which the premises do seem to imply the conclusions. However, I dunno about the law of the excluded middle: either one ought to kill walruses or one ought not to kill walruses. Seems to me it could plausibly be both, or neither. (The trouble here is, of course, that practical reasoning is a matter of weighing evils against each other, and "ought" is too blunt a term to be any use.)
PS Hilary Putnam argued, in "Mathematics without Foundations," that object talk can quite generally be re-expressed as modal talk and vice versa. I imagine this applies to the debate over cognitivism, and potentially settles it. Now where's my prize.