First five timepoints were discarded from the analysis to allow for MR equilibration. Slicing time correction was applied, and all functional MRI data were motion corrected using rigid body registration using SPM software (http://www.fil.ion.ucl.ac.uk/spm/). Next, linear trends were removed from each voxel time series. We also removed nuisance variance in the data by regressing out mean time series from ventricular CSF, white matter, as well as six head motion parameters. Importantly, spatial smoothing and spatial normalization were not performed. Spatial smoothing negates the advantages gained by smaller voxels sizes. In addition, spatial smoothing is employed in traditional general linear model based activation analysis to satisfy the assumptions of random field theory. We did not perform those kinds of analysis and hence found it unnecessary to spatially smooth the data. Next, the Freesurfer analysis pipeline enables individual-specific cortical parcellation from which we extracted the time series used in the analysis. Therefore, we found that spatially normalizing the data into a common space and incurring the costs of blurring and registration errors associated with such a procedure was unnecessary and may be counter-productive for the small voxel size we had and the type of analysis we planned.
