[gumby]

> > true evidence needs to be verifiable

> Scientific evidence needs to be repeatable.

I think the GP comment had it right.

Consider I make some cosmological prediction based on Bob’ observational data (“stars with this spectrum mostly contain elements X and Y”). It does little good to repeat my analysis, which is entirely done on a piece of paper or in a computer program.

But Alice could say, “Well if that’s true, then this other thing would have to be true too” and go check that.

That’s the difference between “verifiable” and “repeatable”.

Sometimes “repeatable” is a sensible form of verification. Last year we spent 9 months trying to reproduce results from an important (to us) paper. Eventually we came up with a reliable process to reproduce the results of the paper — the hypothesis was correct — but it looks like the author just saw some signal a few times, and didn’t really demonstrate the principle they were trying to validate. We couldn’t use the same process described in the paper.

Since we can now get the result whenever we want we consider the theory valid.