Although functional magnetic resonance imaging (FMRI) has arguably become the most ubiquitously used imaging modality, questions remain about the reproducibility of the observed patterns of activation and the acquisition time required to achieve statistically significant and reproducible maps. In the current study, we investigated the reliability of activation in Broca's area, on both a voxel-wise and region of interest level, in response to the covert generation of a single word at 4 T. We also assessed the effects of different parametric (P < 0.01; P < 0.005; P < 0.001) and spatial thresholds (25%, 50% and 75%) on the reproducibility of activation within our region of interest and other randomly selected areas of the brain. We report that the inter-trial consistency of activation within Broca's area for a single trial design using multi-echo EPI is roughly equivalent to previous studies that averaged across a much larger number of trials. However, reliability estimates varied dramatically (approximately 55%) depending on the different parametric and spatial criteria thresholds that were applied to the data. These results show that increased sensitivity at high field strength can be used to reduce the time needed to localize functional activation patterns, which is beneficial for clinical studies such as pre-surgical mapping. Additional benefits of single trial designs, such as the ability to immediately assess for extraneous cognitive processes, are also discussed.