This was something that we argued up and down on a thread about science being able to derive morality.
I and a couple of others (Kevin?) argued that no, you have to assume a moral value first, but you can use reason to show whether your values are consistent or if such and such an action furthers that value, but NOT that the value is a scientific one.
Some others, including Tassman, I believe, started waving around Sam Harris's book, The Moral Landscape and claiming that the book had explained how moral values were too scientifically derived. Yet, those people showed no evidence of understanding the book and much of what they argued showed evidence that they hadn't even read it, often claiming that Harris was arguing from an evolutionary point of view, when that view is explicitly repudiated within the pages of the book.
Why? Because behaviour derived from evolution is merely that which has, on balance, been able to shuffle our genes forward through the eons, but unless keeping genes alive is the moral bedrock then it doesn't help out much to tell us whether, say, murder, rape, theft, gluttony, infidelity and lying is bad, if such behaviour helps you pass on your genes.
That said, I think that while science does not tell us what is moral, we can at least test our moral values against our intuitions.