Given a scarcity of journal space, what is the socially optimal  rule for determining whether an empirical finding should be published? We show that if the goal of publication is to inform the public about a policy-relevant state of the world, then one should publish extreme  results. For specific policy objectives, the optimal rule may take the form of a one- or two-sided test comparing the point estimate to the prior mean. Dynamic considerations may additionally justify the publication of precise null results. If one insists on a rule for which standard inference remains valid conditional on publication, however, the publication rule must not select on the study's findings (but may still select on the study's design).