If it is for all sciences, what methods can be undertaken to ensure that journals print more repeat experiments and failures to replicate instead of only focusing on the "sexy" new ideas?
It's most certainly a problem for all sciences, but there is at least one method to help which is already being employed in some fields. There was a
study publish recently in PLOS One that illustates this well. It compared studies into treatment for cardiovascular disease funded by the NHLBI. Prior to the year 2000, 57% showed a positive effect of the treatment studied. Since 2000, 8% did.
The difference? In 2000 compulsory pre-registration of trials was introduced. This means that negative results are now published, and the aims of the trial had to be clearly specified in advance, to prevent posthoc playing around with statistics to make it look like you discovered something else worthwhile.
Likelihood of Null Effects of Large NHLBI Clinical Trials Has Increased over Time | PLOS ONE
™ Version 4.2
(1)
