There are a number of answers to this question.
1. It has become clear that access to large datasets enhances discovery in ways not possible with smaller datasets, but it is almost impossible for any individual laboratory to collect the massive amounts of data needed for such large-scale analyses.
2. Sharing data allows others to ask questions using your data that you might not have envisioned, thus enabling the greatly possible use of those data.
3. Most fMRI research is funded by public money, which implies that the public has a right to see that those data are used for the greatest possible good.
The most successful current data sharing projects (such as BrainMap, Neurosynth and SUMS-DB) are focused primarily on sharing of processed data such as coordinates or statistical maps. The OpenfMRI project differs in that it provides the basis for sharing of complete raw fMRI datasets (however, it has since evolved to accept raw MRI dataset accompanied by preprocessed data).
Previously, the major project that supported sharing of full fMRI datasets was the fMRI Data Center. It currently has 107 datasets available by request, but has not accepted submission of additional datasets since 2007. In addition, the data are not available online, but instead must be requested for delivery on physical media.
The projects most similar to OpenfMRI in character are the OASIS project (which has shared a large amount of structural MRI data) and the 1000 Functional Connectomes project (which has shared more than 1000 subjects worth of resting-state fMRI data), and NITRC (nitrc.org) (the Neuroimaging Informatics Tools and Resources Clearinghouse) which provides a general purpose image-sharing functionality.
Prior to distribution, the data must be de-identified, which means that any links between the identity of the subject and their data must be destroyed. This includes removal of facial structures from high-resolution structural MRI data.
If you are interested in sharing data through OpenfMRI, you should first receive approval from your IRB for sharing of de-identified data. Most IRB's do not consider data sharing itself to fall under the definition of "human subjects research", but interested users should request such a determination from their IRB before sharing data. The OpenfMRI curators can assist in the deidentification of the data.
Unless otherwise noted, this database and its contents are made available under the Public Domain Dedication and License v1.0 whose full text can be found at: http://www.opendatacommons.org/licenses/pddl/1.0/. We hope that all users will follow the ODC Attribution/Share-Alike Community Norms; in particular, while not legally required, we hope that all users of the data will acknowledge the OpenfMRI project in any publications.
In some cases, it may be necessary to license particular datasets in a more restrictive fashion; the license for each dataset is listed at the bottom of the data page.
In some cases, a dataset must be revised, either to fix errors or add additional information. Minor changes will be noted in the revision history file that accompanies each dataset. In the case that major changes are made (e.g., replacement of primary data files), the revised data file will be given a new accession number and be made available alongside the original dataset, in order to allow persistent access to each version.
If for some reason authors request that a dataset be retracted, the dataset will not be removed from the archive. Instead, it will be accompanied with a notice of retraction (explaining the reasons for retraction) and labeled clearly as having been retracted. This policy ensures that datasets will remain persistent over time, which is required by many journals.
Discussions about OpenfMRI datasets are encouraged. We recommend comments and discussion topics about OpenfMRI datasets be directed to the NeuroStars web site. Please tag any discussion topics with the tags openfmri and dsXXX accession number.