> One could argue that science is about falsifying hypothesis - and terby narrowing in on not just if something works, but how.

Testing the "if" is science - but difficult without a framework helping understand "how". Testing only if, in a vacuum, isn't enough - and it is a tool on the path to understanding "how". This allows such things as using statistical models (if one has an idea of distribution, for example) to help refine the "if" tests, and narrow down exactly what your "if" experiment tested.

I don't think you can have science without both, and I think defining a body of theory is the more important aspect.