John Mark Reynolds' response has helped me to clarify where I think he and I are disagreeing on the torture question.
JMR defends his view based on his argument that torture is worse than killing. Of course, I can easily concede that torture can be worse than killing. But I can't accept that it always is. In 8th grade, a fellow student of mine used to give me wet willies at pretty regular intervals throughout the day. I consider that torture of a very weak sort. It was evil for him to do so, and how I responded was also evil. I got so fed up that I kicked him in the family jewels (or maybe I kneed him). I also consider that to be a kind of torture. Neither is worse than killing someone. Both inflicted pain of a particularly excessive sort. Both involve using someone as a mere means to an end, in his case to take delight in someone else's pain, in my case to satisfy my desire for revenge. Both are wrong, but both are less wrong than killing.
A police officer might cause severe psychological pain by lying to someone in a way that leads to a confession. A police officer might cause physical pain, as long as it's not severe and permanent. Both are perfectly legal in law enforcement, although they are not for military interrogators, who aren't allowed to lie or even touch a suspected terrorist, which was why the Bush Administration wanted more allowed for CIA interrogators, since standard military interrogations were ineffective against al Qaeda. I count such things as mild torture, and they don't seem all that wrong to me, even if one might argue that they are wrong. They certainly aren't worse than killing the suspect.
The key issue is that torture comes in degrees. Killing does too but not in the same sense. What makes an act of killing worse is how you do it, why you do it, and so on. But killing is killing. Torture can be fairly weakly torture, or it can
be pretty awful torture. Killing can be worse because it also involves torture, but the killing itself is not worse. It's the torture that adds to the badness of the killing. Killing is all-or-nothing, and torture comes in degrees in a way that
begs for an analysis that's more complex than simple right and wrong. It's at least a spectrum from not as bad to extremely bad, and it may well be a spectrum from morally required to significantly evil.
I think Mark Olson's question is helpful, so I'll repeat here my comment on his post:
If you only consider consequentialist principles, you can't get an absolute prohibition on anything except the principle that we should seek the best consequences. So to get a moral ban on all torture, there better be some deontological principles at stake. The question is whether those deontological principles themselves are absolutist.[cross-posted at Evangel]
I happen to think the only deontological principle that is absolutist is the moral claim that we ought to give due honor to God and follow him in whatever ways are best for doing so. There are many ethical principles beyond that, and most of them apply most of the time. Some of them apply almost all the time and would require crazy hypotheticals to find exceptions (or very weak cases with moral principles about matters that involve vague concepts occurring along a spectrum, such as consent and coercion, harm, or what someone's motivation and desires are in doing an act).
But to answer your question, I think the deontological principle behind JMR's opposition to torture is his principle that it's always wrong to coerce someone, a principle I've questioned. He doesn't think it's always wrong to cause pain or to cause pain that someone remembers. He doesn't think it's always wrong to cause harm, even permanent harm, knowing full well that one is doing so. He does think it's wrong to cause harm for the sake of causing harm without some higher purpose, but he doesn't think such a higher purpose can be merely getting information that will lead to better consequences when the causing of harm is done to violate someone's ability to consent to giving up that information. So it's mainly an issue of consent to choose to speak when one wants to and to refrain from giving information when one doesn't.
As I've said along the way, I think the problem with that argument is that consent and coercion come along a spectrum, and weaker versions of coercion undermining consent can be morally correct under the right sorts of circumstances. When the consequences increase in their badness, avoiding them might require undermining consent to a stronger degree than is normally moral. That's why I think torture isn't in principle wrong. JMR has a more absolutist view about that principle. We both take it to be deontological, because neither of us thinks slightly better consequences for a serious undermining consent are enough to justify it. But he takes it to be absolutist, whereas I think there could be circumstances where undermining someone's consent via coercion to a great degree can be morally all right, as long as those consequences are extremely serious.
So I'm not sure the disagreement is really meta-ethical. It's more on the level of normative ethics, I think.