[colorint]

>Once the errors are taken out of the system, observation beats theory hands-down.

But how do you go about distinguishing signal from noise? That's ultimately the reason data say far less, on their own, than they seem to: because interpretation of data is at least as important as how you collect them, but interpretation brings in all the gooey things people really want to wish away. Or put more bluntly, observation can't "beat" theory, because theory is the way you decide which observations to make, how to carry out those observations, and how to understand the products of the observation. Theory and observation are inseparable; in a sense, all observation is at once theoretical, and all theory is at once observational.