To clarify: Facebook was shown as the default login option, with a site-specific login available with a Javascript toggle. My test compared which login option was the default. I don't remember the specific numbers (had a lot of traffic though). Google Web Optimizer had given a 95%+ indication that the FB login was performing poorly. So I'd acknowledge issues with the test, but these numbers were convincing enough that I made the site-specific login the default from that point forward. What I did not test is whether having FB login as an option helps or hurts total signups. I suspect that this only helps but can't give any numbers.

Clearly there are lots of caveats, and the results of any experiment are going to be informed by your site, audience, and so on.