Purpose: Functional magnetic resonance imaging (fMRI) is a well-established, non-invasive method for pre-operative mapping of speech-eloquent areas. This investigation tests three simple paradigms to evaluate speech lateralisation and visualisation of speech-eloquent areas. Materials and Methods: 14 healthy volunteers and 16 brain tumour patients were given three tasks: to enumerate months in the correct order (EM), to generate verbs fitting to a given noun (GV) and to generate words fitting to a given alphabetic character (GW). We used a blocked design with 80 measurements which consisted of 4 intervals of speech activation alternating with relaxation periods. The data were analysed on the basis of the general linear model using Brainvoyager®. The activated clusters in the inferior frontal (Broca) and the posterior temporal (Wernicke) cortex were analysed and the laterality indices calculated. Results: In both groups the paradigms GV and GW activated the Broca's area very robustly. Visualisation of the Wernicke's area was best achieved by the paradigm GV. The paradigm EM did not reliably stimulate either the frontal or the temporal cortex. Frontal lateralisation was best determined by GW and GV, temporal lateralisation by GV. Conclusion: The paradigms GV and GW visualise two essential aspects of speech processing: semantic word processing and word production. In a clinical setting with brain tumour patients, both, GV and GW can be used to visualise frontal and temporal speech areas, and to determine speech dominance.