Replications Everywhere

The debate about the replication crisis in the experimental sciences is based on two key claims, namely 1) that researchers rarely replicate existing data and 2) that if they attempt to do so they more often than not fail. These claims have led to some serious soul-searching within the scientific community, the majority of the debate focusing on two issues: a) how can researchers be encouraged to perform more replications and b) how can it be ensured that fewer irreproducible data are created in the first place? These are not idle debates because they could have serious consequences for the way in which research is conducted and funded. Some proposals made in response to the crisis go as far as suggesting that “blue-sky” basic research should be severely restricted in favor of research that is directly tied to practical outcomes. Here, I do not want to contribute to ongoing science policy debates, but instead question the very foundation on which these debates are built. In particular, I will argue that claim 1) is wrong because there is more replication going on in the experimental sciences than usually assumed. These replications, however, are completely ignored in the current debate because the analytic framework used cannot account for them. This also has implications for claim 2), because these additional replications are normally successful, making it likely that current estimates of the failure rate for replications (ranging from 50% to 80 %) are too high. This suggests that there might be less of a crisis than some analysts claim. This of course does not mean that issues such as better quality controls or the reporting of methods do not have to be addressed. Clearly, there is a lot that can be further improved in the experimental sciences. But what it means is that the doomsdaypicture of a profound crisis that drives current calls for reform becomes less convincing. It also means that more focus should be put on getting as complete a picture of the status quo as