In case anyone wants to read that, it's here at PLos One.

I can see some room for error here but I'd actually be curious if anyone else notices what I did or not, with a quick note that while they did explicitly mention removing everything from facial hair to piercings, and adjusting luminance, I didn't see any mention of makeup or cosmetics, maybe I missed it and even if not, maybe they just just forgot to mention it, but that would be one of the possible errors I picked out. I also wish they had included bisexuals too, rather than explicitly removing them, but I can appreciate the desire to limit parameters with a small sample, low hundreds.

Jumping past all of that, what bothers me is that if you remove all the hair, piercings, make up, etc you've really only got 5 things left.

Skin color: I've never heard of indicators homosexuality was more prevalent by ethnicity though I suppose it could be.

Bone Structure: Same as color, only more so, I really can't see bone structure, facial symmetry, eye/ear spacing and position, etc being related to orientation... I could easily believe people think they are subconsciously but that shouldn't matter in the experiment.

Complexion: A person can actually effect that and I could at least see an outside chance gay men are more likely to use skin cream then straight men and maybe vice-versa with women, but that's seem a stretch.

Facial Expression: I could kinda see a decent chance homosexuals are more likely to wear guarded or reserved expressions, or something of a similar variety, but still that wouldn't seem too probable.

Body Fat: Another stretch, but like facial structure I could easily believe people thought there was, subconsciously, but while body fat is much more under someone's control than bone structure I can't think of anything obvious that would make it orientation based, but hormone level do effect body fat IIRC so maybe, again it would seem a stretch.

So, I'm not actually criticizing the experiment, especially since I really just skimmed it, but I didn't see any hypothesis presented as to why this should be, and unless someone can see something wrong with the 5 parameters I mentioned or think of another, there just doesn't seem a method that could cause it, and it makes me lean to thinking something major is being overlooked in the flawed experiment sense.