For fMRI, images are acquired multiple times to describe a time series, while the person being studied performs tasks (cognitive, sensory, motor, etc.) to systematically vary their neuronal activity so that the di˜erences in the images corresponding to di˜erent neural states can be reliably detected.  e great challenge for fMRI is to determine which image changes were due to the neural function and which may have been caused by random noise, physiological motion,

movement of the person being studied, or subtle changes in the MRI system itself (such as due to heating, vibration, electrical power supply ¥uctuations, etc.).  e basic approach underlying the fMRI method results in both its e˜ectiveness and its key limitations and technical challenges.  e primary limitation of fMRI is that it can only show di˜erences in neural function between states.  e design of the reference or baseline state is therefore just as important as the design of the stimulus state. One of the key technical challenges is to understand the relationship between neural activity and the MR image intensity. A second key challenge is to determine which signal changes are truly related to neural functions and which are not.  e most important developments with fMRI have been to respond to these challenges, and these are discussed in the sections that follow.