• Synthesis

• Feedback Cycles

• Shared responsibility and sustainable cooperation

Hypothesizing after results are known ('HARKing') is generally seen as producing less reliable results compared to the orthodox scientific protocol, where the hypothesis is specified in advance. Nevertheless, HARKing remains an important part of scientific practice, and cannot be avoided in important areas such as meta-analyses and research synthesis more generally. In this article I advocate for an account of HARKing which locates this difference in reliability in the manner in which researchers (the 'reporting mechanism') choose to report a particular statistical analysis as evidence in favour of a particular hypothesis. I show that such selective reporting can strongly influence the reliability of the results. This account of HARKing also yields information about situations in which HARKing is less dangerous. I use this to advocate for several methodological and interpretational recommendations to make meta-analytic results more reliable.

Under what circumstances is HARKing less reliable than prediction, and by how much?

• -

-

—

—