Computer simulations and event-related functional MRI (ER-fMRI) experiments were performed to investigate the effects of single-trial averaging and the corresponding contrast-to-noise ratio (CNR) on the minimal resolvable hemodynamic timing difference between brain areas. Three ER-fMRI sessions with temporally delayed (250, 500 and 1,000 ms) visual stimulations between two hemifields, each with 70 repeated single trials, were examined on two subjects. From the computer simulation, the temporal resolution improved as the CNR increased, which reached 500 and 100 ms for CNRs of 1.55 and 6.44, respectively. In the ER-fMRI experiments, the measured CNR increased as more single trials were averaged. The detectability of temporal differences was positively correlated (P<.05) with the CNR in all sessions for one subject but only in the 1,000-ms session for the other subject. Temporal resolution of 1,000 ms was achieved when more than 10 trials were averaged. The 500- and 250-ms delays might be differentiable when more than 20 trials were averaged, but the results were subject-dependent. This study demonstrated that the CNR could be significantly improved by single-trial averaging, which led to an improved temporal resolution of ER-fMRI. Temporal resolution in the range of hundreds of milliseconds was subject-dependent, which might be attributed to the intrinsic spatial variations in the timing of the blood oxygenation level-dependent (BOLD) response.