Failed Replications Show Cracks in Scientific Process

the Neurology Advisor take:

Researchers from the University of Amsterdam have reignited a long debate over the soundness of scientific findings by trying and failing to replicate the results of several studies that claimed correlations between brain structure and behavior.  

Wouter Boekel, of the University of Amsterdam, and colleagues analyzed and tried to replicate the findings of five studies that reported 17 separate positive structural brain-behavior correlations. The researchers conducted structural MRI and DWI scans of 36 students, and then administered questionnaires and psychological tests identical to those administered in the five studies.

Boekel and his team pre-registered the study’s methodology and statistical analysis so as to not change the parameters of the study in order to create positive results—a much debated step in the scientific process.

The researchers found that none of the 17 brain-behavior correlations were significant during replication using Bayesian statistics. However, the researchers said that this doesn’t mean that the correlations are false, but only that a single replication cannot be conclusive when confirming or refuting a claim.

The researchers hope that their findings further support the argument for more preregistered science.