[I admit there is nothing new below. Kai Neilsen made the same basic moves before me. But I think the synopsis is helpful. Also, I refer to god as ‘he’ because I’m a dumbass and a flake and can’t think of an elegant, non-sexist rendering.]
A common objection to atheism is that, by removing god, one removes the “grounds” for morality. Without a divine legislator, humans are free to do and think whatever they want. Below, I attempt to show not only that (a) god is not needed to ground morals, but (b) he cannot be a ground for them.
God and morals: The wrong kind of atheist response
Atheists have been quick to respond that they are, in fact, no less moral than anyone else. And while I am loathe to equate “imprisoned” with “immoral,” I expect many theists do, so it may be relevant that the prison population boasts about the same percentage of declared atheists as the broader population.
Chris Hitchens gives an interesting twist on this defense. In The Portable Atheist, he issues a challenge to believers: “Name me an ethical statement made or an action performed by a believer that could not have been made or performed by a non-believer.”
Hitchens’ point is that there is no logical reason why atheists could not behave just as theists. Assuming this is true, however, it probably doesn’t answer the theist’s main concerns: Supposing there is a divine legislator whose will provides the reason (and the only reason) for ethical behavior. Sure, atheists could make any statements and perform any actions they want all day long—just as theists could say and do unethical things, in spite of god’s commands. The point is that they could only do so inconsistently.
The believer argues that, without a belief in god, there is no rational grounds for behaving morally; an atheist can do it, but he cannot justify it.
Fleshing out the theist challenge
The heart of the theist’s worry is over arbitrariness in morals. Unless morals are imported from outside humanity, each human is left free to define the good for himself, or to dispose of moral categories altogether. To be able to say that some choices are wrong requires a collective “touchstone” or measure against which the choices can be evaluated. We can still argue about what is good or bad; but as C.S. Lewis pointed out, that we can argue at all presupposes there is some real standard, independent of the arguers, to be argued over.
But how does god solve this problem? For he is in precisely the same existential position as we, it seems. He has no “touchstone” outside himself. His choices are just as arbitrary. We would do no worse to elect one of our species to legislate—my uncle Ron, perhaps; his communications would be more direct, no doubt, and would avoid messy controversies about the legislator’s existence. If there are problems with “Lord Ron,” invoking god does not so much solve them as push them back one level.
Believers would respond—they would have to—that god is more than a useful place-holder, an expedient “tie-breaker” in moral disputes. He does not so much ‘pick’ the good as he (including his will) just is good. Only if we assume god is perfectly good—good completely and at all times—can we know that whatever actions he wills for us are good.
Two options for theists
Option 1: Morals by divine command
Still, how god’s will relates to the good is unclear. To paraphrase Plato’s Euthyphro dialogue: Is it good because god wills is, or does he will it because it is good?
On the one hand, we could simply define “good” as “whatever god wills.” Then we know for certain that god’s commands are good because “good” simply means “whatever god commands.” But this just thrusts us back upon the King Ron problem. It doesn’t solve the problem of arbitrariness in morals. God might have willed us to eat babies (he could change his mind and will this tomorrow) and it could not but be “good” to do so.
Some theists bite the bullet and accept this unattractive consequence—at least when asked. But it is unlikely they could ever embrace it in practice. Defining “good” as “what god wills” makes nonsense of much of what believers want to say about morality.
They can never, for example, speak of god’s having good (or any other) reasons for what he wills. It cannot be that in his wisdom god chooses this because it is good. This would imply he is referencing some standard of the good outside himself, which determines his choices—when we have already defined him as the sole standard.
Not to mention most theists (traditional Christians, for one) would be put off by the implication that an ‘almighty’ god should be “determined” or “limited” by some force outside himself in this way.
Likewise, every believer is taught that she always ought to do what god wills. (Certainly, the Bible demands this.) But this too becomes impossible. It invokes the same “outside standard” our definitions have forsworn. When “what we ought to do” and “what god wills” have the same meaning, the most one can say is “I ought to do what I ought to do,” or “god wills what god wills.” This bleeds our moral imperatives of substance, rendering them vacuously tautological.
Option 2: God as “substantially” but not logically good
Still, believers need to say that god’s commands are always good if they are to retain the idea that god is the ground of morals. They just need some non-tautological way of doing so. In short: We need to say that what god wills is always, necessarily good, but it is not good by definition. For lack of a better analogy, think of a dog which is not logically brown—dogs can be all kinds of colors, and this dog could have been another—but is still, in fact, always brown.
But this brings up a new epistemological quandary: How could we know god is perfectly good? There was no problem knowing he was good when we were simply defining him that way; we knew this in the same way we know unmarried males are always bachelors, or dogs (of whatever color) are always canines.
One could claim to know god is good because of direct acquaintance with him or his works. His goodness is known in the in the same way we know our a friend’s penchant for jokes or cooking style.
But to know whether god is good requires a prior and independent understanding of the concept of goodness. We have to know what “good” means before we can look and see if god and his works exhibit this quality. His goodness is not just a cognition but a re-cognition.
This explodes the theist’s argument: If we can know what is good before knowing god, this knowledge cannot depend upon him. We cannot “get our morals from god” because we need the morals to know whether he is in any position to give them out.
While the “grounding” of morals is an interesting problem, it is a problem for both theists and atheists alike. Postulating divinity doesn’t get us closer to solving it. Even if there were a god, we would have to get a sense of right and wrong from a source outside him. Thus, the fact that we do possess a sense of right and wrong cannot be an argument for the existence of god.
 Hitchens’ approach exemplifies the problems I have with most of today’s “leading lights” (“brights”?) of atheism. They speak as though the only bad consequences theism could produce are observable: Religion fuels violence, divides people, slows progress on stem cell research, etc. This prejudice leads him to cast the ethical aspect of the debate as one of how we get people to do right stuff.
 The theists have a point: Living an authentically moral life requires more than the right behaviors. It requires having the right reasons behind them. Having good reasons is what makes moral actions moral in the first place: If I trip and accidentally cushion a baby’s fall from a balcony, I am hardly to be praised for his rescue. Indeed, I could have been on my way to throw a different baby off a different balcony; maybe I only tripped because I was wearing socks to better sneak into the apartment. Even if I meant to save the baby, but only in a crass bid for media exposure, it ceases to be a moral act. If, however, I place myself in the baby’s way deliberately, I’m a hero, and deserve all the praise I get. The behavior is the same; only the reasons distinguish them.
 There are other problems with knowing god’s goodness that exceed the scope of this piece. Note, god is not just good but perfectly good. It is unclear how one could be “acquainted” with any “perfect” quality. Maybe we can know god has been good up to now, but “perfect goodness” projects this quality of goodness into an infinite future—while our data is ever limited to the past and present. But this for another post.