A solution to this could be to integrate replication studies in the evaluation process of scientific papers, e.g., as we suggest in our #PeerReplication model: doi.org/10.1038/s443..., and to be very careful when building on scientific discoveries that havn't yet been replicated independently!