DescentBB

http://www.its.caltech.edu/~dg/HowScien.pdf

By David Goodstein , B.S., M.S., Ph.D., is Vice Provost, Professor of Physics and Applied Physics, and the
Frank J. Gilloon Distinguished Teaching and Service Professor, California Institute of Technology, Pasadena, California.

Best 16 page article you are likely to read this year.

OR...

Most important article you skipped because you're too lazy to care.

it was OK, but not \"the best\". (You've been posting a lot about \"the best\" writings of various sorts, and so far, none of them have turned out to be all that great.)

-----

Science, at its core, is a loop you can start at any point:

Observe -> Theorize -> Test -> [REPEAT]

You observe something, try to describe the whole of the evidence in a consistant way, and then test your description against the evidence (by looking at the evidence in more detail, gathering new evidence, etc.) Or, you're handed a theory from someone else, you test it against available evidence, make new observations, and modify the theory. In any case, you continue to repeat the process -- you continue to build a more robust theory based on analysis of more pieces of evidence.

Here, \"evidence\" can refer to basically anything that should be taken into consideration when evaluating the theory. Very careful chemistry experiments are a form of evidence; observations about the way a certain group of people behave are another. The main difference between types of evidence is how solidly convincing they are -- when someone can exactly replicate your experiment, they're more likely to be convinced than if they have to make their own observations that might not exactly match yours. One of the strongest types of evidence is when a theory allows one to make predictions, and those predictions are consistantly confirmed.

There are two reasons Peer Review is important: first, it means you have a lot more sets of eyes doing the \"test\" step based on a wider collection of observations. This greatly reduces the chance of missing an obvious sign that a theory is broken. Second, a good scientific theory should be reasonably convincing -- perhaps not \"this is the right theory\", but at the very least \"this might work; this guy might be on to something.\" If those who are peer-reviewing your theory consistantly fail to see any value in it, it's either broken or underdeveloped; either way, you need to spend more time working on it.