Rob J

Not true. If I was to repeat the red hat experiment twice with a probability of (1/10 million)squared the whole world would take notice (if it wasn't such an absurd example and so unbelievable). If it was a reasonably authoritative body doing the experiment it might even change the scientific cannon immediately. It would be much harder to get this reaction with a retrospective analysis. Prospective analysis is always more potent than retrospective analysis. A lot of pseudoscience (psychics etc) is reliant on this. People will talk about coincidences in their past and claim psychic ability. They don't predict winning lottery numbers. Btw, by analogy with your plant example, the mechanism causing the red hat to increase chance of winning the lottery pre-exists but not the the phenomenon of winning the lottery.

The mechanisms pre-existed but the phenomenon of the plant growing better didn't. An example of the difference between prospective/ retrospective analysis: If I make a prediction that wearing a red hat will win me the lottery tomorrow and I win the lottery, that's pretty impressive. Plus, if it's a genuine correlation I can repeat it again and again. A 1/10 million lottery repeated 3 times has a (1/10 million)cubed probability If I look back at the evidence and find that a lot of people with red hats have won the lottery, even if I can show it's statistically significant, it's not so impressive. To increase the significance all you can do increase the size of your dataset.

Sure, and it also won't be exactly 50/50.

Could you clarify what you are saying here here? In the most rigorous form of scientific method a theory makes a prediction that you can then test. Are you using the term 'prediction' to refer to predicting what pre-existing 'evidence' you are going to find? In other words, an example of the usual definition of prediction in scientific method is tossing a coin and predicting 50/50 (ie theory->prediction->test prediction) Are you using 'prediction' to refer to, for instance, what kind of fossil you are going to find? This is better than nothing but ideally science should make prospective predictions not retrospective - that makes is easier to introduce bias because you will tend to find and notice the evidence you want, and ignore the rest. But someone has to decide what 'the most likely explanation' is. If you can come up with a fantastic predictive experiment you can 'prove' something in one paper. Most things involve slowly building up the mass of evidence until the scientific community accepts it as the most likely explanation. There is a probability level that can a paper published. There is no fixed probablility level for when a theory becomes part of scientific cannon, that comes from majority consensus. I read the blog. However: The difference is we have seen the cat and haven't seen the milk fairy, so yes the milk fairy involves inventing/creating a new entity. However both evolution and God have to be 'created' to explain the data.

You can't observe macroevolution.

Sometimes science is straightforward - you can just measure data to support your case eg toss a coin and count it and you can show head/tails is roughly 50/50. Sometimes it is about having a theory that explains the data and slowly building on this case so that the scientific community accepts it as the most likely explanation. If your case it strong enough it almost becomes accepted as 'fact'. An example is evolution - the fossil record, mutations in genes and observation of microbial evolution all support it as the most likely explanation. However Creationists would say that God fits the same purpose, a single unifying theory that explains all the observations. Philosophically how can you differentiate between these two?