Functional mapping of the human brain has made tremendous progress in the past years thanks to new technical developments. Imaging methods are now available; they allow to study brain functions with high spatial and temporal resolution. Single photon emission computer tomography (SPECT), positron emission tomography (PET), functional magnetic resonance imaging (fMRI) and high resolution electro- and magnetoencephalography (EEG and MEG) are currently intensively applied techniques to functional studies, each one having specific properties concerning spatial and temporal resolution. The success of these methods in basic neuroscience research has led to the demand for applying them to clinical questions. Diseases of the central nervous system that lead to brain dysfunction can be ideally explored using these techniques. Of particular importance are those diseases in which a focal neuronal dysfunction is the primary cause and where surgical resection of this focus might be the cure. This is often the case for epilepsy, where a discrete primary focus might exist from which pathological rhythms evolve and propagate throughout the brain, leading to seizures that severely handicap the patient. Surgical resection of the primary focus is only possible if the focus can be exactly localized and adequately separated from functionally important areas. This is where these new functional imaging tools become important. The use of SPECT and PET for focus localization has been most extensively studied and their specificity and sensitivity are intensively discussed. In the last few years functional MRI has evolved as a new interesting tool in epileptic focus localization. The most important limitation of these techniques, however, is the temporal resolution. Since epileptic activity can propagate very fast, several hyper- or hypoactive regions are seen in the images and primary areas cannot be distinguished from regions of propagation. The only methods that have sufficient temporal resolution to follow neuronal activity in real time are the electrophysiological measures, i.e. the EEG and the MEG. Localization of the sources in the brain that produced a given surface electromagnetic field has become possible through algorithms that solve the so-called "inverse problem". Several different algorithms exist and many groups begun to apply them to epileptic data with the aim to localize the focus of the pathological electrical discharges. This review article discusses the use of distributed EEG source localization procedures in the presurgical evaluation of patients with intractable focal epilepsy. In contrast to equivalent dipole models, distributed localization methods do not localize one active point in the brain but rather assume extended active areas, which is generally the case in epileptic activity. The methods shown here are based on linear numerical methods and are therefore less prone to errors when working with scattered solution spaces such as the one defined by anatomical constraints. Solutions constraint to the gray matter determined in the individual MRI are shown here. We illustrate three methods to increase the spatial resolution of the source localization procedures: One is to increase the number of recording channels to more than 100, the second to use linear methods of high precision to detect focal sources (EPIFOCUS), and the third to combine EEG source localization with EEG-triggered functional magnetic resonance imaging. The importance of EEG source localization for the interpretation of fMRI data will be particularly discussed in view of the important difference of the temporal resolution by the two methods. The localization methods can be applied to interictal as well as to ictal activity. In case of analysis of ictal EEG we propose to use full scalp frequency analysis to determine the time period of seizure onset and to localize the sources of the initial dominant frequency.