fMRI retinotopic mapping provides detailed information about the correspondence between the visual field and its cortical representation in the individual subject. Besides providing for the possibility of unambiguously localizing functional imaging data with respect to the functional architecture of the visual system, it is a powerful tool for the investigation of retinotopic properties of visual areas in the healthy and impaired brain. fMRI retinotopic mapping differs conceptually from a more traditional volume-based, block-type, or event-related analysis, in terms of both the surface-based analysis of the data and the phase-encoded paradigm. Several methodological works related to fMRI retinotopic mapping have been published. However, a detailed description of all the methods involved, discussing the steps from stimulus design to the processing of phase data on the surface, is still missing. We describe here step by step our methodology for the complete processing chain. Besides reusing methods proposed by other researchers in the field, we introduce original ones: improved stimuli for the mapping of polar angle retinotopy, a method of assigning volume-based functional data to the surface, and a way of weighting phase information optimally to account for the SNR obtained locally. To assess the robustness of these methods we present a study performed on three subjects, demonstrating the reproducibility of the delineation of low order visual areas.