This paper argues that some of the discussion around meta-scientific issues can be viewed as an argument over different “meta-hypotheses” – assumptions made about how different hypotheses in a scientific literature relate to each other. I argue that, currently, such meta-hypotheses are typically left unstated except in methodological papers and that the consequence of this practice is that it is hard to determine what can be learned from a direct replication study. I argue in favor of a procedure dubbed the “limited homogeneity assumption” – assuming very little heterogeneity of effect sizes when a literature is initiated but switching to an assumption of heterogeneity once an initial finding has been successfully replicated in a direct replication study. Until that has happened, we do not allow the literature to proceed to a mature stage. This procedure will elevate the scientific status of direct replication studies in science. Following this procedure, a well-designed direct replication study is a means of falsifying an overall claim in an early phase of a literature and thus sets up a hurdle against the canonization of false facts in the behavioral sciences.

本文认为，关于元科学问题的一些讨论可以被视为对不同“元假设”的争论，这些假设是关于科学文献中不同假设之间如何相互联系的假设。我认为，目前，除了方法论论文外，这些元假设通常都未被阐明，这种做法的结果是很难确定可以从直接复制研究中学到什么。我主张采用一种被称为“有限的同质性假设”的程序-假设在启动文献时效应大小的异质性很小，但是一旦在直接复制研究中成功复制了最初的发现，便切换到异质性的假设。在此之前，我们不允许文献进入成熟阶段。该程序将提高直接复制研究在科学中的科学地位。按照这一程序，精心设计的直接复制研究是伪造文献早期阶段的总体主张的一种手段，从而为行为科学中的错误事实的规范化设置了障碍。