Echo-planar imaging with gradient echo allows whole-brain images to be rapidly acquired. However, its main limitation is that magnetic field inhomogeneities in regions of the brain close to bone and air-filled sinuses result in reduced signal-to-noise ratio and signal loss. In particular this is a problem for imaging the temporal lobes, and can therefore affect the results of some language-related studies. Decreasing the echo time (TE) increases the signal-to-noise ratio and reduces the amount of signal loss in susceptible regions. In this study we investigate the TE dependence of BOLD (blood oxygenation level-dependent) contrast and, in particular, how it is influenced in regions with susceptibility artifacts. We use a dual echo-time sequence to compare brain activations measured with two different TEs, TE = 40 ms and TE = 27 ms. The paradigm involves comparing famous faces to scrambled faces, a low-level control condition. It was chosen because famous faces have been repeatedly shown to activate the fusiform gyri and anterior temporal lobes in both PET and fMRI. Our results show that it is possible to detect robust activations at a lower TE in brain regions not affected by susceptibility artifacts (i.e., fusiform gyri), allowing for faster scanning times. However, although the amount of signal loss is reduced at the lower TE, this does not appear to be sufficient to recover the BOLD signal in regions affected by susceptibility artifacts (i.e., anterior temporal lobes).