Software-based data analysis is not limited to neuroimaging, and the fact today is that neuroscientists from different academic disciplines spend an increasing amount of time writing software to process their experimental data. We would wager that almost all scientific data are ultimately processed with computers software. The practical consequence of this is that the quality of the science produced relies on the quality of the software written (Dubois, 2005). The success of digital data analysis is made possible not just by acquiring high-quality data and sometimes by using sophisticated numerical and mathematical methods, but critically it is made possible by using correct implementations of methods. The MNE-Python project is developed and maintained to help provide the best quality in terms of accuracy, efficiency, and readability. In order to preserve analysis accuracy, the development process requires the writing of unit and regression tests (so-called test-driven development) that ensure that the software is installed and functioning correctly, yielding results that match those previously obtained from many different users and machines. This testing framework currently covers about 86% of the lines of MNE-Python
