One reason for that is said to be that journals have a bias towards wanting to publish 'positive' results - that is, reports on experimental testing of a hypothesis that result in the researcher being able to say 'Yay! My idea works out', rather than 'Well, it was an interesting idea, but my experiments showed no evidence that it was correct." Particularly in a field like psychology (It's a bit less likely to happen in more hard-edged sciences like physics) this puts pressure on researchers to find something - anything - in their results that can be said to confirm some idea they may or may not have had before doing the experiment. At best, this kind of publication adds meaningless noise to the field, and at worst it is actually misleading.
So do we say. "Hmm, looks like science is not a reliable way of discovering true things after all"? Nope, because science is a huge and complex social institution that is both self-critical and creatively self-correcting: look at this new idea used by the Royal Society (the world's oldest scientific society):
Registered Reports
A Registered Report is a form of journal article in which methods and proposed analyses are pre-registered and peer-reviewed prior to research being conducted (stage 1). High quality protocols are then provisionally accepted for publication before data collection commences. The format is open to attempts of replication as well as novel studies. Once the study is completed, the author will finish the article including results and discussion sections (stage 2). This will be appraised by the reviewers, and provided necessary conditions are met, publication is virtually guaranteed.
The main benefits of this two-stage approach are:
Once the methods and proposed analyses are provisionally accepted, the journal will commit to publishing the results regardless of the outcome, provided the final study conforms to the initially approved proposal and meets all quality checks. This means that publication bias is reduced as negative results will not prevent publication.
Peer review of the research proposal provides an opportunity for the authors to receive constructive critical feedback that may help them to fine-tune the study design prior to conducting the experiment.
This process can help reduce researcher bias.
This process may enhance the credibility of the work.
Neat: you describe the experiments you plan to do, and what research questions you are trying to answer. If the journal thinks your work is worth doing, the journal agrees to publish your results even if the result is, in effect, 'We did not get any positive results'. You get a publication (important to your career), and other researchers now know that the experiments you tried led to a dead end, so they need to think of something else. And the temptation to select only the 'interesting' results from your study is much diminished - because you have already set on record with a third party how your experiment will be done.
Science has dealt with a major problem by admitting it has one, and then creating an ingenious method of self-correction.
Can anyone give me an example of religion doing that?