It's been a while since I learned the scientific method per se, but I remember something about "repeatable, controlled experiments". There are many, many questions in life just can't be studied in a repeatable, controlled experiment. Macroeconomic policy comes to mind. (You might reason, "okay, I'm going to make two Earths, and in the US on planet 1 I'm going to implement policy A, then in the US on planet 2 I'm going to implement policy B, then I'm going to travel back in time to the present day so we have identical initial conditions and implement the better policy.") Anything single-shot. Anything involving a counter-factual ("if Hitler had never been born, would there have been a second world war?"). Anything where confounding variables cannot be eliminated or controlled. I could continue but I think the point is probably clear. :)