The manuscript of Pernet and Poline, "Improving functional magnetic resonance imaging reproducibility", is clearly written and addresses a timely issue. Reproducible Research is gaining momentum (e.g., the recent book "Implementing Reproducible Research" edited by Stodden, Leisch and Peng whose chapters are available on-line at: https://osf.io/s9tya/wiki/home/) and many scientists attracted to the idea are starting wondering how to go from a "nice principle" to its actual implementation. This manuscript addresses this practical aspect targeting the fMRI community but in way that makes it a profitable reading for a much wider audience.

In their short introduction the authors start by setting and locating the "reproduction problem". They then rightly point to the computer science community that for its own reasons, mainly the production of reliable and easy to maintain software, has developed the tools needed for reproducible research implementation. They go on and explain that if the fMRI practitioners are unlikely to have received the formal training of computer scientists, the barrier for using the required tools is getting lower and lower.

In their next section "Reproducible neuroimaging in 5 easy steps", the authors "decompose" the production of an fMRI paper in 5 steps, discussing for each the procedures to implement in order reach reproducibility; the software, data repository, etc, available and insisting on the benefits for the community beyond strict reproducibility, that is protocol, data and code reuse.

In the short "Tips for more reliable and replicable code: lessons from bioinformatics" section the authors advocate for the use of script based analysis (as opposed to point and click) and point to useful resources to improve the quality / readability / maintainability of the scripts. I can only agree with them: using scripts is the easiest way to make analysis reproducible and making the community understand that point is one of the key challenges.

In the "Discussion" section they come back to the benefits of sharing (data, code, etc) and then discuss the interesting idea of "preregistration" with examples and pointers to relevant web site.

The "Annex 1 – list of websites mentioned in the article, which can be used for sharing" section contains a useful compilation of the web sites mentioned in the manuscript, with a short description of each.

This is a well written, concise and very useful manuscript.

Discretionary revisions

In subsection "Organize and share data and meta-data", the link to the "W3C Provenance Group" is missing.

I think citing the book "Implementing Reproducible Research" should help readers wanting to go further. Level of interest An article of outstanding merit and interest in its field Quality of written English Acceptable Declaration of competing interests No competing interest.

Author response:

http://www.gigasciencejournal.com/imedia/4514699461599493_comment.pdf