Functional Magnetic Resonance Imaging (fMRI) allows the exploration of neural activity in the brain at a spatial resolution of a few millimetres and a temporal resolution of several seconds. The combination of functional and anatomical images, acquired in the same listener, allows for a direct relation of physiological processes to the underlying anatomical structures in the brain.

The principle of MRI is based on radio frequency excitation and relaxation of nuclear magnetic moments that have been aligned with a strong external magnetic field. Functional MRI is usually based on the narrow magnetic resonance of proton spins. Spatial encoding is achieved using additional gradient coils that restrict RF excitation to predefined slices through the tissue. The spatial resolution and signal-to-noise ratio are directly linked to the strength of the external magnetic field. For imaging human subjects, field strengths from 1.5 Tesla up to 9 Tesla are used.

With fMRI, a secondary, metabolic response is observed rather than the primary neural activation. The actual signal change of the measured NMR signal is based on the change of the oxygenation level of blood due to metabolic processes that are triggered by the neural activity. These changes result in a slight change of the relaxation times in activated regions of the brain and are the basis for an image contrast between different regions. As a consequence, the observed signal change has a comparatively long latency of several seconds relative to stimulus onset.

Since the observed NMR signal changes in activated regions of the brain are very small (in the order of 1%), the analysis of fMRI data is heavily based on repeated measures and statistical models, that test for significant differences between conditions. A wide-spread approach also employed in all fMRI studies summarised in this thesis is the method of statistical parametric mapping, based on the theory of Gaussian random fields.