Functional magnetic resonance imaging (fMRI) gained a lot of interest in medical and human research in the last years. FMRI is a noninvasive method used to study human brain functions by localizing activated brain areas. There are a lot of interesting questions in neurobiology, one of these is the processing of learning-related processes in the human brain and how these processes can be described and analyzed. To analyze fMRI data different hypothesis-based and data-based methods can be used. The Independent Component Analysis (ICA) is an information-theoretic statistical and computational technique used to identify hidden factors of observed multivariate data. The mathematical background of ICA is investigated in this thesis and relations to other methods like Principal Component Analysis (PCA) are elaborated. The advantages of ICA in comparison to classical methods for analyzing fMRI data under the aspect of learning-related processes are investigated in real fMRI data as well as in simulations studies. Thereby dynamic changes in the fMRI time series are systematically analyzed and described.