Great blog on statistics here – recent post suggests, however, that not many would be interested in a journal of null results. Commenters pointed out that there are several journals like this already. We DO need a repository for these – especially in medicine where it's actually fairly useful to know if an intervention (particularly one that we've been doing, or has been sold to us) does not work. I agree with one suggestion, though – maybe not a whole journal article, but some methodological details (a bad study proving no benefit is not valuable, but a good study is…) and the results might be sufficient, with a really good indexing system. You'd also need the site/journal to confer at least a little benefit for the authors going to the trouble of writing it up.