In a recent episode of the Big Picture Science podcast, there was an interview with Stuart Firestein (chair of the Columbia University Biology Department) in which he discussed his book Ignorance: How It Drives Science. In particular, he mentioned that for the most part he dislikes hypotheses---they tend to pigeonhole people's idea of what "good data" is, so that if when unexpected happens, it's often thrown out as "bad data".
Is this a problem in physics? If so, how much of a problem is it, and what, if anything, do physicists do to attempt to mitigate it?
My instinct tells me that this is a problem, albeit possibly not as much as in so-called "soft" fields such as psychology and sociology where eliminating variables and bias is much more difficult (at least I assume it's much more difficult).