Simultaneous multi-slice (SMS) echo-planar imaging (EPI) has been developed to speed up acquisition of whole-brain fMRI data⁽¹⁾. It is the aim of the current study to evaluate possible gains in statistical sensitivity of SMS EPI with short TR as well as the potential of available software packages to account for enhanced serial autocorrelations in fMRI data with high temporal resolution. Determining optimal imaging parameters as well as valid analysis strategies for detection of single event-related responses is relevant for a broad variety of fMRI studies (e.g. oddball paradigms, sustained attention tasks) and could also be used in psychiatric disorders (e.g. responses to threat-related stimuli in anxiety disorders) as well as brain diseases which are characterized by short and distinct alterations of neural function, such as interictal epileptiform discharges.15 healthy volunteers participated in this fMRI study comprising seven imaging runs after providing informed consent. To define the region of interest (ROI) within the visual cortex, a short block paradigm (duration: 3 minutes) with blocks of flickering checkerboards and blocks with a fixation cross (duration of each block: 20 seconds) was employed. The following six imaging runs (duration: 10 minutes per run) were carried out with six different MB factors (1, 2, 4, 5, 8, and 10) resulting in TRs ranging form 2.64 s to 0.26 s. The fMRI images were acquired at a voxel size: 3 × 3 × 3 mm³, with a 3 T scanner (Siemens TIM TRIO, Erlangen, Germany) equipped with a 32-chnnel head coil. An event-related design with brief checkerboard stimulation periods (duration: 500 ms) was used in these runs. The order of these runs was counterbalanced across subjects. Since we aim to develop an optimized imaging protocol for detection of spikes in epileptic patients in future studies, the design (temporal occurrence of events) was based on the spike discharge pattern of an epileptic patient examined in our department and contained 17 single events with a mean inter stimulus interval (ISI) of 34.1 s ranging from 24.9 s to 50.3 s. We calculated whole-brain extra square of sum (ESS) images to examine where physiological noise is correlated with BOLD effects. In addition we also systematically evaluated the impact of slice-time correction (SPM 8), and the effect of various autoregressive (AR) orders (FMRISTAT⁽²⁾).As expected, the block-design (checkerboard > fixation cross) yielded widespread activation in the visual cortex (t(14) = 16.9 at activation peak, MNI coordinates: x = 9, y = -85 , z = -5, k = 1255 voxels). This area was defined as ROI for subsequent analysis of event-related responses obtained at different MB factors. The difference between t-values with and without slice time correction further decreased with higher temporal resolution (Figure 1). Additional inclusion of physiological noise regressors had only a very small effect on resulting t values within the ROI. Only a very small part of our ROI (Figure 2, white area) overlapped with voxels showing an effect for respiration. In Figure 3 the resulting t-values dependent on the used MB factor and the order of the AR model are shown. At shorter TRs (≤ 0.33 s), increasing the order of the AR model by one, resulted in a statistically significant (p< 0.05, Bonferoni corrected) lower t scores (5 - 17%) up to AR order 3.In this study, we examined the impact of colored noise (i.e. physiological noise due to heart beat and respiration as well as neural noise arising from spontaneous neural activity), temporal resolution, and employed method for correction of serial autocorrelations on statistical inference (t-values) using event-related fMRI in combination with SMS EPI with various MB factors resulting in TRs ranging from 2.64 s to 0.26 s. Statistical power of fMRI time series increased with sampling frequency arguing for improved sensitivity of fast fMRI acquisition methods in event-related designs. There seem to be limits of possible gains, however, as highest t-values were observed for a TR of 0.33s (MB-8) which then decreased for data obtained with a TR of 0.26s (MB-10) due to signal loss caused by incomplete T1 relaxation and noise enhancement associated with higher acceleration factors which also limits gains of SMS-EPI⁽³⁾. TRs < 500 ms yielded increased t-scores which suggest that higher order AR models are necessary for correction of colored noise in such data.