This study aims to quantify the mean change of the effective transverse relaxation time T2⁎ in active brain regions of human volunteers at field strengths of B0=3T and 7T. Besides the mono-exponential signal decay model an extended model is tested that considers mesoscopic field gradients across imaging voxels. Both models are checked for cross-talk and correlations between the parameters. A visual checkerboard-stimulation experiment with pause and stimulation periods of 50s and six repetitions was performed on healthy volunteers. Eleven contrasts were acquired in about 1.47s/1.43s at 3T/7T using a segmented multi-contrast echo-planar imaging (EPI) sequence. Average BOLD-signal time courses were calculated in a multi-step (non-)linear least-squares process. Baseline T2⁎ values of 37.72ms/24.99ms (47.34ms/33.71ms) with stimulus-correlated changes ∆T2⁎of 1.32ms/0.74ms (1.99ms/1.43ms) resulted from the mono-exponential (extended) model for 3T/7T. A dependence of those values on the initial intensity S0 was observed. Stimulus-correlated changes of S0 in the order of 1% were measured at both field strengths. The mono-exponential model was found to be less prone to instabilities in the regression of both parameters. Signal alterations due to inflow were observed. Measured relaxation times agree with values from literature using repetitive stimulation. A strong dependence of the measured relaxation times on the inflow-related model parameter was found for both models. The extended model is applicable to dynamic neurofunctional measurements, but is currently limited due to the low number of contrasts acquired.