Continuous raw data are stored in instances of the Raw class. MNE-Python supports reading raw data from various file formats e.g., BTI/4D, KIT, EDF, Biosemi BDF and BrainVision EEG. Other formats such as eXimia or CTF can be converted to FIF files using tools available in the MNE-C package, also available at http://martinos.org/mne. The Neo project (Garcia et al., under review) implements readers in Python for micromed and elan files, which can facilitate the use of these formats with MNE-Python. The FIF file format allows organization of any type of information into a multi-leaved tree structure of elements known as tags. It is at the core of the MNE-Python package which favored the development of highly optimized reading and writing routines for this format. It offers for example the ability to read data from disk only when needed. This access-on-demand principle can also be inherited by other classes that build upon Raw (such as Epochs and Evoked, below), which offers the possibility to process data with a very limited memory usage. Typical processing steps at this stage include filtering, noise
