Based on systematic in vivo studies analysed by fuzzy clustering, we prove the complex dependence of functional magnetic resonance imaging (fMRI) signal changes on echo time (TE) (ranging from 42 ms up to 160 ms) in the human visual cortex at 1.5 Tesla. We obtain a steady increase of signal enhancement with increasing TE, except at TE = 130 ms where the signal increase is significantly less than at 100 and 160 ms, respectively, caused by signal dephasing of spins in the vascular environment of the cortical region imaged. A comparison with model calculations reveals that vessels with a diameter of about 0.5 to 1 mm may be the main source of gradient-recalled echo fMRI signal changes at 1.5 Tesla at the given spatial resolution. In addition to conventional correlation analysis, fuzzy cluster analysis has been applied to evaluate fMRI data sets. Our results also indicate that, despite the similar temporal pattern of the functional response, it is possible to differentiate between areas which show higher signal enhancement (cluster 1, higher blood volume fraction) and lower signal enhancement (cluster 2, lower blood volume fraction), reflecting the different vascular environment. Therefore, fuzzy cluster analysis may help to extract functional information from activated areas closer to the actual neuronal activation.