What are the roadblocks to successful scientific replications? | Pete Etchells

Replication is the bedrock of science. But what happens when a scientist can't - or won't - share the experimental materials that allow it?

It's a simple question, but one that's essential to science: if I repeat an experiment, how reliably can I get the same result? But it's a question that you can't ask without the right materials. It's like baking a cake - if you're not given enough instructions in the recipe, or told what sort of ingredients you're supposed to be using, you might end up with a carrot cake when you thought you were making a Battenberg.

For psychological research, replication is big business at the moment. In August this year, the Reproducibility Project - a groundbreaking attempt to systematically assess the reliability of published psychological research findings - delivered a grim result. Of 100 experiments that were replicated, the original findings were only reproduced in just 36% of cases. Some news outlets saw this as an opportunity to take a dig at the entire discipline, as if this was somehow conclusive evidence that psychology wasn't a real science. But low replication rates are an issue that extend beyond psychology - cancer biology is facing the difficult reality of irreproducible results, and a recent analysis of 67 economics papers found that even if the original authors helped out, only 49% of results were reproducible. Other outlets reported much more thoughtfully on the findings from the Reproducibility Project though, and highlighted that the real take-home message was aimed squarely at psychologists: there's more work to be done. We need more replication studies, and they need to become an acceptable, respected, and ingrained part of psychological research life.