Rosenthal addresses the 'file drawer problem', a questionable research practice where only studies that showed the desired result would be published and all other studies would land in the 'file drawer' and thus unknown to the scientific community.
In the extreme case, this could mean that, if a specific effect did not exist in reality, the 5% of studies that (due to statistical error allowed) find this effect get published and discussed as if the effect were true, whereas 95% of studies do not (and rightly so) find the effect, but are tucked away in a file drawer. This problem hinders scientific progress, as new studies would build on old, but false, effects.
Rosenthal introduces a way to assess the size of the file drawer problem, the tolerance to future null results: calculating the number of studies with null results that would have to be in a file drawer before the published studies on this effect would be called into question.