Hmmm. My knee-jerk reaction was “a lot!” but then I started thinking more… and really, the process could very easily be the same (hypothesis, testing, etc) except when it comes to the evaluation of the data. It depends, how do you evaluate a musical experiment? When I think of classic scientific experiments they seek to prove a constant… that is, there is something they want to prove that exists 100% of the time. Everything has to be reproducible, yielding the same results. For example, if you want to prove alcoholism has a genetic component… well, you can use humans but usually this is too tricky and it’s easier to use a similar genetic model (which, sadly for rats, works out well for scientists). In music, you may not be able to do testing on similar genetic models (guess it depends on what you’re trying to prove) and instead have to involve us more tricky humans.
I guess in the end scientific experiments are easier, more cut and dry because of the controlled conditions they can be carried out in. Musical experiments I would guess are probably pretty difficult to conduct (perhaps like psychology experiments which are notoriously difficult because of the human factor). There would be qualitative data vs quantitative. And reproduction would likely be challenging.