Better living through transparency: Improving the reproducibility of fMRI results through comprehensive methods reporting

Recent studies suggest that a greater proportion of published scientific findings than expected cannot be replicated. The field of functional neuroimaging research is no exception to this trend, with estimates of false positive results ranging from 10 % to 40 %. While false positive results in neuroimaging studies stem from a variety of causes, incomplete methodological reporting is perhaps the most obvious: Most published reports of neuroimaging studies provide ambiguous or incomplete descriptions of their methods and results. If neuroimaging researchers do not report methods and results in adequate detail, independent scientists can neither check their work for errors nor accurately replicate their efforts. Thus, I argue that comprehensive methods reporting is essential for reproducible research. I recommend three strategies for improving transparency and reproducibility in neuroimaging research: improving natural language descriptions of research protocols; sharing source code for data collection and analysis; and sharing formal, machine-readable representations of methods and results. Last, I discuss the technological and cultural barriers to implementing these recommendations and suggest steps toward overcoming those barriers.