Only half of the published research findings in the disciplines such as economics, psychology and education can be replicated by others according to new findings reported in Nature by Dr Dermot Lynott, Lead of the TROPIC project and Chair of the Irish Reproducibility Network, in collaboration with international colleagues.

The study, the largest of its kinds, was conducted as part of the SCORE project (Systematizing Confidence in Open Research and Evidence), where researchers replicated studies from 164 papers across the social and behavioural sciences. A paper was considered replicable if a new analysis of different data relevant to the same question yielded the same overall result, with the overall successful replication rate of 49.3%. Dermot’s role in the project was to work with the replication teams and scrutinise their protocols – “We wanted to be sure that what they planned to do really matched the original study so that if a result did or did not replicate, we could say with confidence that it was a fair test.” He adds that, “Researchers need to be beyond reproach – that means putting everything on the table – all of our data, our materials, our statistical code and our computational methods – so that others can check our work and build on it. Science is messy and uncertain by nature – we can’t change that, but we can change how honest and open we are about the process, and that’s the best way to protect public trust in the long run.”

An interesting finding from another paper in the SCORE project was that reproducibility of studies was easier and more likely to succeed when research practices were open and transparent.

For more information on activities of the Irish Reproducibility Network, visit the Irish Reproducibility Network website.

The Nature paper is available here.

A summary of the findings of the SCORE project is available to read on Science.

Image credit: Piquant. Image CC:BY.