I've mentioned it before but I think that the between/within groups distinction is a little misleading. It needs to be broken up by prevalence of specific attributes in each group. I don't mean to criticize you or expect you to have that kind of data off the top of your head though. I just think that a large enough sample to cut through the noise of individual preference might reveal some interesting, but subtle, differences. Given the current genetic and developmental hypothesises for sexual orientation I think it would make sense if there was a great deal of overlap but not a complete overlap of preferences.
I don't see how it's misleading.
There is undoubtedly overlap between groups in preference when it comes to partner selection and attraction in general, but I honestly think it's such an individualized and personal thing, that "breaking it up into specific attributes" does violence to the original phenomena one is trying to measure. Reductionism often misses out on the larger picture in my opinion, and this is a perfect example. My original point is that, on the one hand, one woman to another or one man to another is going to exhibit on average a far greater difference than a particular attribute compared across groups. That's just basic statistics. However in another sense, the relationships and interplay between those attributes---their context---are just as important as the attributes themselves, and so I think "the noise of individual preference" is precisely what is important.
Also I'm a little uncomfortable with the idea of applying purely biological models of sexual attraction to groups based on a factor that isn't purely biological, ie. gender.
Edited by MuppetFace - 9/14/12 at 12:06pm