Disclaimer: this post may be relevant only for social sciences/psychology people. I found a nice thread on the Judgment and decision making (JDM) mailing list on the need for replications.
Lots of good posts on an interesting discussion. The mainstream view is that we simply don’t run enough replications because they are harder to get published. This leads to studies showing that replications are actually very hard, with only a small percentage (about 40% in the social sciences) being successful. Robyn Dawes seems to thing that replications are overrated:
the “real” scientists do is to futch around until they get it “right.” The multiple study requirement just adds “first and second and third” studies, thereby wasting space and time.