[Lockyy]

Being familiar with the client introduces another area of potential bias, removing that allows us to compare the specific variable being tested for.

I'm not entirely sure why the big voices thing is important. It's not like they were saying all the voices had low favourability or that what were perceived as gay voices had low favourability, they were demonstrating a difference between the two groups based on that one variable. Whether they were all high up the rankings or all low down the rankings is irrelevant, you just need them to be all within the same band. It's the difference that is what is being tested for.