[duwease]

IMO, the change has to come from the grant financiers, and a case can be made that they get a greater more bang for their buck by adding requirements such as pre-registration, open data publishing (so long as they don't want to keep the data proprietary, of course), and publishing guaranteed regardless of outcome. The fruits of what they purchased can be of more use to the field in this manner, and even they themselves may find they spend less money if the 'failures' are documented in such a way that they don't inadvertently repeatedly fund them due to the fact that the initial results were hidden away in a filing cabinet.

I think the same case could be made for placing more emphasis on funding replication -- instead of thinking 'we learned it, why revisit it?', you're checking if it's actually valuable information worth spending more on, or just a fluke.