I’ve heard it said that the only ‘real’ science is experiment-based – things like palaeontology, astronomy & so on can’t get it right because they are ‘observational’, looking back into the past, & so can’t generate repeatable data. It’s usually said as the preface to special pleading for some sort of supernatural influence on the world.
Now Ken has a great post on this rather pointless distinction, over at Open Parachute.