A field that definitely has a problem with replication is Computer Human Interaction. There are a lot of user studies in that field and you basically never see a study done twice. The setup of the studies usually doesn't even allow it to be repeated as it hinges on some proprietary software written for that very study that is not released to the public.