Functional MRI relies on the ability to detect the changes in NMR signal produced in discrete regions of cortex in response to specific activating stimul, which are believed to reflect changes in local blood flow, volume and oxygenation. The underlying mechanism includes susceptibility contrast effects based on changes in deoxyhemoglobin in blood. We have used functional MRI to map cortical responses non-invasively in several studies, including sensory and motor activation, and studies of visual processing, language, auditory and spatial working memory and attention. The results demonstrate the power of this new technique. However, there remain several unanswered questions regarding its optimal mode of use, the tissue and technical factors that are important in determining the signal changes detected, and their interpretation. Artifacts in computed activation maps may be produced by tissue motion or other physiological noise, and these limit the sensitivity of detecting small changes. A critical feature of current paradigms for detecting activation is the method of data analysis, which is interrelated with the nature of the task and imaging method used. Different methods of analysis have been exploited in our studies to date.