This thesis presents electroencephalography (EEG) brain imaging by covering topics as empirical evaluation of source confusion, probabilistic inverse methods, and source analysis performed on infant EEG data. In terms of source confusion we inspect how current sources within the brain may be confused with each other as noise is present in the EEG recordings. Moreover, we examine how errors in the forward model affect the source confusion. The primary aim of this thesis is to provide sharper EEG brain images by improving current inverse methods. In this relation we focus the attention on two topics in EEG source reconstruction, namely, the forward progation model (describing the mapping from the current sources within the brain to the sensors at the scalp) and the temporal patterns present in the EEG. As forward models may suffer from a number of errors including the geometrical representation of the human head, the tissue conductivity distribution, and electrode positions, we propose an algorithm which consider forward model uncertainties. Bayesian graphical models provide a powerful means of incorporating prior assumptions that narrow the solution space and lead to tractable posterior distributions over the unknown sources given the observed data. Here, we propose a hierarchical Bayesian model that attempts to minimize the influence of uncertainties associated with the forward model on the source estimates. Similarly, we develop a hierarchical spatio-temporal Bayesian model that accommodates the principled computation of sparse spatial and smooth temporal EEG source reconstructions consistent with neurophysiological assumptions in a variety of event-related imaging paradigms.