This is a really great point. In fact, _any_ survey with self-selecting participants is highly susceptible to selection bias. What surveys like this are good for is measuring the opinions of a group of people, weighted by degree of explicit interest in the subject. Most people just use the default light-backgrounded-schemes, and don't even think about it. The silent majority.

However, this survey is useful in that it reveals that there is indeed a passionate core of users who prefer dark backgrounds. The survey author seems to understand all this, as his conclusion is that it is definitely worth adding a dark background option, but that it may or may not be the default.