What is the rational basis for banning same-sex marriage?

My colleague in the Political Science department, Howard Schweber, emails a question about same-sex marriage that I know some of my readers will be good at answering. (And he likes the idea of my reprinting this to get answers from people who actually believe in banning same-sex marriage.)

Assuming that rational basis scrutiny is the appropriate level of review (which is far from clear in the majority opinions in Romer and Lawrence, both of which ducked the question), what is the strongest case that can be made for a legitimate state interest in restricting marriage to opposite-sex couples? I take it as given that if the statutes are constitutional, the fear of judicial intervention is an adequate explanation for turning them into constitutional amendments, so it's the original statutes that I am thinking of.

[One answer is] that majoritarian moral preference, standing alone, is an adequate justification for legislation. One reason I have doubts about that principle is its implications for free speech, actually, but let's leave that aside.

Ann made a more empirical suggestion -- I hope I am accurately recounting it as follows:

1. We prefer heterosexual couples to homosexual couples
2. There exist some number of persons, however small, who enter into heterosexual marriages because homosexual marriages are unavailable.

Note that I didn't say that I personally have these preferences. I'm trying to say what a court might find to be a rational basis if it were considering the constitutionality of the state constitutional amendment. As I've said many times on the blog, I support same-sex marriage, and I do not disparage gay relationships. I'm simply saying that a court might find rationality in the expression of special respect for the traditional relationship and that this respect -- with additional benefits and protections -- will encourage more people to form these relationships. I'm not saying this is a good thing to believe, just that it is one belief that is at least rational. I am assuming there are some people who are influenced by social pressure to form traditional male-female families who would, with sufficient social approval, chose a same-sex relationship.

Back to Howard:

One question is empirical; does a state need any evidence to demonstrate the existence of this class of persons?

The other question, it seems to me, is the basis for the preference. Avoiding the appeal to morality, one gets to the policy arguments (better for children, etc.) These seem awfully weak, and indeed analytically incoherent. [One might] say moral preference is enough, but that doesn't necessarily solve the problem. Moral preference between classes of persons, after all, are definitely not a permissible basis for legislation -- it has to be moral preference about conduct (Scalia says "lifestyle,") right? In other words, we are talking about sex.

So we come down to a moral preference for the kinds of sexual conduct that would be engaged in by the class of persons -- perhaps they are bisexual, for example -- who would enter into heterosexual marriages today but would not have done so if the alternative of homosexual marriage were available.

The problem there is selectivity: we don't use any similar moral preferences in determining who may marry in any other context. In the debate the other night I proposed the following indicia for concluding that expressions of a purported state interest are a pretext for animus:

novelty - an argument of a kind we have never heard before
selectivity -- a principle that is not applied in any other context
targeting -- a principle that is not applied to any other class of citizens
extremism -- a principle that if applied consistently would yield obviously unacceptable results.

Answers?